![\"Open](\"https://colab.research.google.com/assets/colab-badge.svg\"){kind=icon}
\n\n# + colab={\"base_uri\": \"https://localhost:8080/\"} id=\"DAtyt3O7Q9-A\" outputId=\"73df475d-5768-4474-fb90-80d5136844cd\"\n# !pip install mysql-connector-python\n\n# + colab={\"base_uri\": \"https://localhost:8080/\"} id=\"XOZpJ43gQ9-G\" outputId=\"23b475ce-f010-493f-9b8f-ad6e23d11031\"\nimport mysql.connector\nimport pandas as pd\n# Define the connection string (assuming a typical localhost setup for demonstration)\nconnection_string = {\n 'user' : 'doadmin',\n 'password' : 'AVNS_sAMuAiwTX_UzEy6cR63',\n 'host' : 'db-ro-mysql-nyc1-25804-do-user-13295890-0.b.db.ondigitalocean.com',\n 'port' : 25060,\n 'database' :'defaultdb'\n}\n\n# Establish the connection\nconnection = mysql.connector.connect(**connection_string)\n\n# Check if connection is established\nif connection.is_connected():\n print(\"Connected to MySQL server\")\n df = pd.read_sql(\"SELECT*FROM ejercicio1\",connection)\n connection.close()\nelse:\n print(\"Failed to connect\")\n\n# + colab={\"base_uri\": \"https://localhost:8080/\", \"height\": 423} id=\"LLvBeyXzTAyd\" outputId=\"762ee42d-202e-4b11-a57c-375d4c0ebb03\"\ndf\ndf.drop('ID_Venta', axis = 1)\n\n# + colab={\"base_uri\": \"https://localhost:8080/\", \"height\": 229} id=\"Vh4RunxQVrzf\" outputId=\"e744ad96-a5a3-410b-aafb-6f8100383b4e\"\ndf.groupby(\"Hora_Dia\").sum()\n\n# + colab={\"base_uri\": \"https://localhost:8080/\", \"height\": 206} id=\"n6FxJNk9XK4p\" outputId=\"43fb2b30-7ca9-4bc5-f071-e85fc68a7cf5\"\ncategorias = ['Electrónica', 'Ropa', 'Alimentos', 'Muebles']\nhora_dia = ['Mañana', 'Tarde', 'Noche']\nn = 100\ndf = pd.DataFrame({\n 'ID_Venta': np.arange(1, n + 1),\n 'Categoria': np.random.choice(categorias, n),\n 'Hora_Dia': np.random.choice(hora_dia, n),\n 'Unidades_Vendidas': np.random.randint(1, 20, n),\n 'Ingresos': np.random.uniform(100, 500, n)\n})\ndf.head()\n","repo_name":"DanAndCastRod/python_DataScience","sub_path":"session4_2_mysql_conect.ipynb","file_name":"session4_2_mysql_conect.ipynb","file_ext":"py","file_size_in_byte":2228,"program_lang":"python","lang":"en","doc_type":"code","stars":0,"dataset":"github-jupyter-script","pt":"41"}
+{"seq_id":"2407080293","text":"# +\nimport os\nimport tarfile\nimport urllib.request\nimport pandas as pd\n\nDOWNLOAD_ROOT = \"https://raw.githubusercontent.com/ageron/handson-ml2/master/\"\nHOUSING_PATH = os.path.join(\"datasets\", \"housing\")\nHOUSING_URL = DOWNLOAD_ROOT + \"datasets/housing/housing.tgz\"\n\ndef fetch_housing_data(housing_url=HOUSING_URL,housing_path=HOUSING_PATH):\n os.makedirs(housing_path,exist_ok=True)\n tgz_path = os.path.join(housing_path,\"housing.tgz\")\n urllib.request.urlretrieve(housing_url, tgz_path)\n housing_tgz = tarfile.open(tgz_path)\n housing_tgz.extractall(path=housing_path)\n housing_tgz.close()\n\n\ndef load_housing_data(housing_path=HOUSING_PATH):\n csv_path = os.path.join(housing_path,\"housing.csv\")\n return pd.read_csv(csv_path)\n\n\n# +\n#running the functions to fetch and load data\nfetch_housing_data()\nhousing = load_housing_data()\n\n#getting the dataframe information:\n\nprint(housing.head())\nprint(housing.info())\nprint(housing.describe())\nprint(housing[\"ocean_proximity\"].value_counts())\n# -\n\n#data viusualization:\nimport matplotlib.pyplot as plt\nhousing.hist(bins=50, figsize=(20,15))\nplt.show()\n\n#spliting the data into training and testing sets:\nfrom sklearn.model_selection import train_test_split\ntrain_set,test_set=train_test_split(housing,test_size=0.25,random_state=42)\n\n#defining the income categories:\nimport numpy as np\nhousing[\"income_cat\"]= pd.cut(housing[\"median_income\"],bins=[0.,1.5,3,4.5,6.,np.inf],labels=[1,2,3,4,5])\nhousing[\"income_cat\"].hist()\nplt.show()\n\n# +\nfrom sklearn.model_selection import StratifiedShuffleSplit\nsplit = StratifiedShuffleSplit(n_splits=1, test_size=0.2, random_state=42)\nfor train_index, test_index in split.split(housing,housing[\"income_cat\"]):\n strat_train_set=housing.loc[train_index]\n strat_test_set=housing.loc[test_index]\n\n#droping the income cat columns of the set:\nfor set_ in (strat_train_set,strat_test_set):\n set_.drop(\"income_cat\", axis=1, inplace=True)\n# -\n\n#exploring the data through visualization:\nhousing=strat_train_set.copy()\nhousing.plot(kind=\"scatter\",x=\"longitude\",y=\"latitude\",alpha=0.1)\n\n#exploring the data through visualization:\nhousing=strat_train_set.copy()\nhousing.plot(kind=\"scatter\",x=\"longitude\",y=\"latitude\",alpha=0.4,s=housing[\"population\"]/100,label=\"population\",figsize=(10,7),c=\"median_house_value\",cmap=plt.get_cmap(\"jet\"))\nplt.legend()\nplt.show()\n\n# +\n#explorign the data through corelation matrix:\ncorr_matrix = housing.corr()\n\nhousing.head()\n\nprint(list(enumerate(housing.keys())))\n\nprint(corr_matrix[\"median_house_value\"].sort_values(ascending=False))\n\nplt.matshow(housing.corr())\nplt.show()\n# -\n\n#explorign the data through correlation plots:\nfrom pandas.plotting import scatter_matrix\nattrbutes=['median_house_value','median_income','total_rooms','housing_median_age']\nscatter_matrix(housing[attrbutes],figsize=(12,8))\n\n\nprint(list(enumerate(housing.keys())))\n\n# +\n#defining new attributes for the dataset:\nhousing[\"rooms_per_household\"]=housing[\"total_rooms\"]/housing[\"households\"]\nhousing[\"bedrooms_per_room\"]=housing[\"total_bedrooms\"]/housing[\"total_rooms\"]\nhousing[\"population_per_household\"]=housing[\"population\"]/housing[\"households\"]\n\ncorr_matrix = housing.corr()\nprint(corr_matrix[\"median_house_value\"].sort_values(ascending=False))\n\nplt.matshow(housing.corr())\nplt.show()\n\n# +\n#preparing the data for machine learning implementation:\n#clean start:\nhousing = strat_train_set.drop(\"median_house_value\",axis=1)\nhousing_labels = strat_train_set[\"median_house_value\"].copy()\n\n#replaceing the NaN values with the calculated mean.\n'''\nmedian = housing[\"total_bedrooms\"].median()\nhousing[\"total_bedrooms\"].fillna(median,inplace=True)\n'''\nfrom sklearn.impute import SimpleImputer\n\nimputer = SimpleImputer(strategy=\"mean\")\n\n#only numerical data:\nhousing_num = housing.drop(\"ocean_proximity\",axis=1)\nimputer.fit(housing_num)\n#creating final traingin data:\nX = imputer.transform(housing_num)\nhousing_tr=pd.DataFrame(X,columns=housing_num.columns,index=housing_num.index)\n\nprint(X)\nprint(housing_tr)\n# -\n\n#transfering categorical variables into numeric description: - machine learning algoritms dont work on text they work on vectors/tensors\nfrom sklearn.preprocessing import OneHotEncoder\ncat_econder = OneHotEncoder()\nhousing_cat=housing[[\"ocean_proximity\"]]\nhousing_cat_1hot = cat_econder.fit_transform(housing_cat)\n#print(housing_cat_1hot)\nprint(housing_cat_1hot.toarray())\nprint(cat_econder.categories_)\n\n#reminder on what is housing data:\nprint(type(housing)) #a DataFrame\nprint(housing.head())\nprint(housing.keys())\nprint(housing.values)\n\n#creating custom transformers:\nfrom sklearn.base import BaseEstimator, TransformerMixin\n#defining indexes of attibutes in the dataset:\nrooms_ix, bedrooms_ix, population_ix,households_ix= 3,4,5,6\n#defining the class that inherits from sklearn transformers to enable adding parameters to the dataset:\nclass CombinedAttibutesAdder(BaseEstimator,TransformerMixin):\n def __init__(self,add_bedrooms_per_room=True):\n self.add_bedrooms_per_room = add_bedrooms_per_room\n def fit(self,X,y=None):\n return self\n def transform(self,X):\n rooms_per_household=X[:,rooms_ix]/X[:,households_ix]\n population_per_household=X[:,population_ix]/X[:,households_ix]\n if self.add_bedrooms_per_room:\n bedrooms_per_room=X[:,bedrooms_ix]/X[:,rooms_ix]\n return np.c_[X,rooms_per_household,population_per_household,bedrooms_per_room]\n else:\n return np.c_[X,rooms_per_household,population_per_household]\n#read up on what np.c_ does! - arrays concatination along axis.\n'''\nnp.c_[np.array([1,2,3]), np.array([4,5,6])]\narray([[1, 4],\n [2, 5],\n [3, 6]])\n'''\nattr_adder=CombinedAttibutesAdder(add_bedrooms_per_room=False)\nhousing_extra_attirbs = attr_adder.transform(housing.values)\n\n# +\n#feature scaling for optimal learning experience :)\n#STANDARDIZATION & MIN-MAX SCALING as two basica algorythms:\n\n#sklearn PIPELINES: - execution of data transfromations in a tipical order:\nfrom sklearn.pipeline import Pipeline\nfrom sklearn.preprocessing import StandardScaler\nfrom sklearn.impute import SimpleImputer\n\nnum_pipeline = Pipeline([\n #imputer allows to clean the data form NANs and replace them with median values:\n (\"imputer\",SimpleImputer(strategy=\"median\")),\n #attribs_adder has been defined in a previous block as a class that allows to extend the attributes:\n (\"attribs_adder\",CombinedAttibutesAdder()),\n #std_scaler standardizes the data:\n (\"std_scaler\",StandardScaler()),\n])\n\n#learning ready data:\nhousing_num_tr = num_pipeline.fit_transform(housing_num)\n\n\n# +\n#Column transformer - allows to perform transformations of a pipeline column by column and distinguish numeriacal and cathegorical data:\nfrom sklearn.compose import ColumnTransformer\n\n\"\"\"\nprint(housing_num)\nprint(list(housing_num))\n\"\"\"\nnum_attribs=list(housing_num)\ncat_attribs=[\"ocean_proximity\"]\n\n#combined pipeline that applies different transformations based on the attibutes of the datafarme:\nfull_pipeline = ColumnTransformer([\n (\"num\",num_pipeline,num_attribs),\n (\"cat\",OneHotEncoder(),cat_attribs),\n])\n\nhousing_prepared=full_pipeline.fit_transform(housing)\nhousing_labels = strat_train_set[\"median_house_value\"].copy()\n# -\n\nprint(housing)\n\n# +\n#choosing the model and training:\n#linear regression model:\n\nfrom sklearn.linear_model import LinearRegression\nlin_reg = LinearRegression()\nlin_reg.fit(housing_prepared,housing_labels)\n#testing the model:\nsome_data = housing.iloc[:5]\nsome_labels = housing_labels.iloc[:5]\nsome_data_prepared = housing_prepared[:5]\n\nprint(\"Predictions: \",lin_reg.predict(some_data_prepared))\nprint(\"Actual Values: \",list(some_labels))\nprint(lin_reg.score(some_data_prepared,some_labels))\n\n#model underfitting\n\n# +\n#choosing the model and training:\n#decision tree model:\n\nfrom sklearn.tree import DecisionTreeRegressor\ntree_reg = DecisionTreeRegressor()\ntree_reg.fit(housing_prepared,housing_labels)\n#testing the model:\nsome_data = housing.iloc[:5]\nsome_labels = housing_labels.iloc[:5]\nsome_data_prepared = housing_prepared[:5]\n\nprint(\"Predictions: \",tree_reg.predict(some_data_prepared))\nprint(\"Actual Values: \",list(some_labels))\nprint(tree_reg.score(some_data_prepared,some_labels))\n\n#model overfitting\n\n# +\n#model cross validation for more precise understanding of performance:\n\nfrom sklearn.model_selection import cross_val_score\n\nscores = cross_val_score(tree_reg,housing_prepared,housing_labels,scoring=\"neg_mean_squared_error\",cv=10)\ntree_rmse_scores = np.sqrt(-scores)\n\n\n# +\ndef display_scores(scores):\n print(\"Scores: \",scores)\n print(\"Mean :\",scores.mean())\n print(\"Standard Deviation :\",scores.std())\n\ndisplay_scores(tree_rmse_scores)\n# -\n\nlin_scores = cross_val_score(lin_reg,housing_prepared,housing_labels,scoring=\"neg_mean_squared_error\",cv=10)\nlin_rmse_scores = np.sqrt(-lin_scores)\n\n\n# +\ndef display_scores(scores):\n print(\"Scores: \",scores)\n print(\"Mean :\",scores.mean())\n print(\"Standard Deviation :\",scores.std())\n\ndisplay_scores(lin_rmse_scores)\n\n# +\nfrom sklearn.ensemble import RandomForestRegressor\nforest_reg = RandomForestRegressor()\nforest_reg.fit(housing_prepared,housing_labels)\n#testing the model:\nsome_data = housing.iloc[:5]\nsome_labels = housing_labels.iloc[:5]\nsome_data_prepared = housing_prepared[:5]\n\nprint(\"Predictions: \",forest_reg.predict(some_data_prepared))\nprint(\"Actual Values: \",list(some_labels))\nprint(tree_reg.score(some_data_prepared,some_labels))\n\nforest_scores = cross_val_score(forest_reg,housing_prepared,housing_labels,scoring=\"neg_mean_squared_error\",cv=10)\nforest_rmse_scores = np.sqrt(-forest_scores)\n\ndef display_scores(scores):\n print(\"Scores: \",scores)\n print(\"Mean :\",scores.mean())\n print(\"Standard Deviation :\",scores.std())\n\ndisplay_scores(forest_rmse_scores)\n\n# +\n#Finetuning the models:\n\n#01.Grid search\n#02.Randomize search\n#03.Ensemble methods\n\nfrom sklearn.model_selection import GridSearchCV\n\nparam_grid = [\n # try 12 (3×4) combinations of hyperparameters\n {'n_estimators': [3, 10, 30], 'max_features': [2, 4, 6, 8]},\n # then try 6 (2×3) combinations with bootstrap set as False\n {'bootstrap': [False], 'n_estimators': [3, 10], 'max_features': [2, 3, 4]},\n ]\n\nforest_reg = RandomForestRegressor(random_state=42)\n# train across 5 folds, that's a total of (12+6)*5=90 rounds of training \ngrid_search = GridSearchCV(forest_reg, param_grid, cv=5,\n scoring='neg_mean_squared_error',\n return_train_score=True)\ngrid_search.fit(housing_prepared, housing_labels)\n\n# -\n\ngrid_search.best_params_\ngrid_search.best_estimator_\n\n# +\ncvres = grid_search.cv_results_\nfor mean_score, params in zip(cvres[\"mean_test_score\"], cvres[\"params\"]):\n print(np.sqrt(-mean_score), params)\n \npd.DataFrame(grid_search.cv_results_)\n\n# +\nfeature_importances = grid_search.best_estimator_.feature_importances_\nfeature_importances\n\nextra_attribs = [\"rooms_per_hhold\", \"pop_per_hhold\", \"bedrooms_per_room\"]\n#cat_encoder = cat_pipeline.named_steps[\"cat_encoder\"] # old solution\ncat_encoder = full_pipeline.named_transformers_[\"cat\"]\ncat_one_hot_attribs = list(cat_encoder.categories_[0])\nattributes = num_attribs + extra_attribs + cat_one_hot_attribs\nsorted(zip(feature_importances, attributes), reverse=True)\n\n# +\nfrom sklearn.metrics import mean_squared_error\n\nfinal_model = grid_search.best_estimator_\nhousing = strat_train_set.drop(\"median_house_value\",axis=1)\nhousing_labels = strat_train_set[\"median_house_value\"].copy()\n\nhousing_prepared=full_pipeline.fit_transform(housing)\n\nhousing_test = strat_test_set.drop(\"median_house_value\",axis=1)\nhousing_test_lables = strat_test_set[\"median_house_value\"].copy()\n\nhousing_test_prepared=full_pipeline.fit_transform(housing)\n\nfinal_predictions = final_model.predict(housing_test_prepared)\n\nfinal_score = final_model.score(housing_test_prepared,housing_labels)\n\nprint(final_score)\n","repo_name":"baczkowskij/Machine-Learning-With-Scikit-Keras-Tensorflow","sub_path":"Chapter_02_End_to_End_Machine_Learning_Project/End_to_End_Machine_Learning_Project.ipynb","file_name":"End_to_End_Machine_Learning_Project.ipynb","file_ext":"py","file_size_in_byte":11907,"program_lang":"python","lang":"en","doc_type":"code","stars":0,"dataset":"github-jupyter-script","pt":"41"}
+{"seq_id":"16041990807","text":"# + [markdown] id=\"view-in-github\" colab_type=\"text\"\n# \n\n# + [markdown] id=\"6LnkP-PWvBN-\"\n# # 7 - Document Classification Using RNN\n\n# + [markdown] id=\"q5flqwAJv99t\"\n# References:\n# - [RNN with keras](https://www.tensorflow.org/guide/keras/rnn)\n# - [Text Classification using RNN](https://www.tensorflow.org/text/tutorials/text_classification_rnn)\n\n# + [markdown] id=\"mJ3vWYxKw9O-\"\n# ## Dataset\n\n# + [markdown] id=\"B-0rpdNnx_Pf\"\n# - Dataset Used : Imdb Movie Review dataset\n# - Dataset Url : https://ai.stanford.edu/~amaas/data/sentiment/aclImdb_v1.tar.gz\n# - Dataset Home Page: https://ai.stanford.edu/~amaas/data/sentiment\n\n# + [markdown] id=\"Z3vhCm5_bdIB\"\n# ## Importing the required modules\n\n# + id=\"Zo86H7NqbhHx\"\nimport tensorflow as tf\nimport numpy as np\nimport matplotlib.pyplot as plt\nimport os\nimport re\nimport shutil\nimport string\n\n# + [markdown] id=\"recoIGKAgIcF\"\n# ## Dataset\n\n# + [markdown] id=\"lt_oMUN-cBb2\"\n# ### Downloading the dataset \n\n# + [markdown] id=\"X5-Dng2ucPtq\"\n# - Dataset: Imdb Large Movie Review Dataset v1.0 \n# - Home Page : https://ai.stanford.edu/~amaas/data/sentiment/\n# - Dataset URL : https://ai.stanford.edu/~amaas/data/sentiment/aclImdb_v1.tar.gz\n\n# + id=\"y5lbithLcRSR\"\nurl = \"https://ai.stanford.edu/~amaas/data/sentiment/aclImdb_v1.tar.gz\"\n\ndataset = tf.keras.utils.get_file(\n \"aclImdb_v1\", url,\n untar=True, cache_dir='.',\n cache_subdir=''\n)\n\ndataset_dir = os.path.join(os.path.dirname(dataset), 'aclImdb')\n\n\n# + [markdown] id=\"ZQqwkhlcc4Cd\"\n# ### Exploring the dataset directory\n\n# + colab={\"base_uri\": \"https://localhost:8080/\"} id=\"WdsJr12zc6Rm\" outputId=\"1277e0f8-d4b7-4725-92f2-aee85ac94e57\"\nos.listdir(dataset_dir)\n\n# + colab={\"base_uri\": \"https://localhost:8080/\"} id=\"esSEyuDCdl5s\" outputId=\"2b64247c-0b84-40a8-84b2-7c8441c96c4b\"\ntrain_dir = os.path.join(dataset_dir,\"train\")\nos.listdir(train_dir)\n\n# + [markdown] id=\"JrV6GAMseC2V\"\n# Here `pos` and `neg` are the folders containing positive and negative reviews.\n#\n#\n\n# + [markdown] id=\"pwTuaZ7MgQi7\"\n# ### Removing unwanted folders from training folder\n\n# + [markdown] id=\"QlVXlhOAgYpT\"\n# `unsup` contains data for unsupervised learning which is not needed for this exercise. so we remove that using `shutil.rmtree()`\n\n# + id=\"K8zcOZdpd7ki\"\nshutil.rmtree(os.path.join(train_dir,\"unsup\"))\n\n# + [markdown] id=\"jZ2d_5-6e6nh\"\n# ### Loading the dataset\n\n# + colab={\"base_uri\": \"https://localhost:8080/\"} id=\"LQk9TCWUfPuz\" outputId=\"2b7eeb40-27e3-4752-fcc4-eb6540a9e03d\"\nbatch_size = 128\nseed = 42\n\nraw_train_ds = tf.keras.utils.text_dataset_from_directory(\n 'aclImdb/train', \n batch_size=batch_size, \n validation_split=0.2, \n subset='training', \n seed=seed)\n\n# + colab={\"base_uri\": \"https://localhost:8080/\"} id=\"GQvx6SI9g1rR\" outputId=\"1296445a-c3db-4b7f-abe2-8c9331069b44\"\nraw_val_ds = tf.keras.utils.text_dataset_from_directory(\n 'aclImdb/train', \n batch_size=batch_size, \n validation_split=0.2, \n subset='validation', \n seed=seed)\n\n# + colab={\"base_uri\": \"https://localhost:8080/\"} id=\"LrrWVRE4g2OB\" outputId=\"fa868dda-13c2-4dc5-e0fb-9e1f1176805a\"\nraw_test_ds = tf.keras.utils.text_dataset_from_directory(\n 'aclImdb/test', \n batch_size=batch_size)\n\n# + id=\"pkkgKzkop62e\"\nAUTOTUNE = tf.data.AUTOTUNE\n\ntrain_ds = raw_train_ds.cache().prefetch(buffer_size=AUTOTUNE)\nval_ds = raw_val_ds.cache().prefetch(buffer_size=AUTOTUNE)\ntest_ds = raw_test_ds.cache().prefetch(buffer_size=AUTOTUNE)\n\n\n# + [markdown] id=\"ofoJ1-yCgDbe\"\n# ## Preprocessing the text\n\n# + [markdown] id=\"XOEK9hx7hmAW\"\n# ### Standardization Function\n\n# + id=\"OVdcJFvuhpj1\"\ndef custom_standardization(input_data):\n lowercase = tf.strings.lower(input_data)\n stripped_html = tf.strings.regex_replace(lowercase, '\n\n# + id=\"Ov26gDy-UXl0\"\nimport numpy as np\nimport keras\nimport sys\nfrom keras.models import Model\nfrom keras.utils import plot_model\nfrom keras.models import load_model\nfrom PIL import Image\nimport matplotlib.pyplot as plt\nfrom random import randint\nimport imageio\nimport argparse\nfrom skimage.util.shape import view_as_blocks\n\n\n\n# + id=\"t-hRvm_tUicH\"\nmodel_hide=load_model(\"/content/drive/MyDrive/models_mix/hide2.h5\",compile=False)\n\n\n# + id=\"GVM1OZbpU5OO\"\ndef normalize_batch(imgs):\n '''Performs channel-wise z-score normalization'''\n\n return (imgs - np.array([0.485, 0.456, 0.406])) /np.array([0.229, 0.224, 0.225])\n\n# Denormalize outputs\ndef denormalize_batch(imgs,should_clip=True):\n imgs= (imgs * np.array([0.229, 0.224, 0.225])) + np.array([0.485, 0.456, 0.406])\n \n if should_clip:\n imgs= np.clip(imgs,0,1)\n return imgs\n\n# Custom block shuffling\ndef shuffle(im, inverse = False):\n \n # Configure block size, rows and columns\n blk_size=56\n rows=np.uint8(img.shape[0]/blk_size)\n cols=np.uint8(img.shape[1]/blk_size)\n\n # Create a block view on image\n img_blks=view_as_blocks(im,block_shape=(blk_size,blk_size,3)).squeeze()\n img_shuff=np.zeros((img.shape[0],img.shape[1],3),dtype=np.uint8)\n\n # Secret key maps\n map={0:2, 1:0, 2:3, 3:1}\n inv_map = {v: k for k, v in map.items()}\n\n # Perform block shuffling\n for i in range(0,rows):\n for j in range(0,cols):\n x,y = i*blk_size, j*blk_size\n if(inverse):\n img_shuff[x:x+blk_size, y:y+blk_size] = img_blks[inv_map[i],inv_map[j]]\n else:\n img_shuff[x:x+blk_size, y:y+blk_size] = img_blks[map[i],map[j]]\n \n return img_shuff\n\n\n# + id=\"lwx-spPlVITZ\"\nsecretin = np.array(Image.open(\"/content/drive/MyDrive/secret mix test/9031426R.png\").convert('RGB').resize((224,224))).reshape(1,224,224,3)/255\ncoverin = np.array(Image.open(\"/content/drive/MyDrive/new cover_test/00114.jpg\").convert('RGB').resize((224,224))).reshape(1,224,224,3)/255\n\n# + id=\"nHz5_X_8XQGe\"\ncoverout=model_hide.predict([normalize_batch(secretin),normalize_batch(coverin)])\n\n# + id=\"ybZTRry9Xei7\"\ncoverout = denormalize_batch(coverout)\ncoverout=np.squeeze(coverout)*255.0\ncoverout=np.uint8(coverout)\n\n# + colab={\"base_uri\": \"https://localhost:8080/\", \"height\": 286} id=\"Byvf6u2mXk4Z\" outputId=\"8cd4d52a-df02-4135-9a7c-739553d5bcc6\"\nimageio.imsave('container3.png',coverout)\nplt.imshow(coverout)\n","repo_name":"Amoghkori/Deep-Learning-Projects","sub_path":"image_hide.ipynb","file_name":"image_hide.ipynb","file_ext":"py","file_size_in_byte":2639,"program_lang":"python","lang":"en","doc_type":"code","stars":0,"dataset":"github-jupyter-script","pt":"44"}
+{"seq_id":"32572094603","text":"# + id=\"lVAK2Z1dNB5H\"\n#SVM Classification\nimport pandas as pd\nimport numpy as np\nfrom sklearn.preprocessing import StandardScaler\nimport matplotlib.pyplot as plt\nimport seaborn as sns\n\n\n# from sklearn import svm\nfrom sklearn.svm import SVC\nfrom sklearn.model_selection import GridSearchCV,RandomizedSearchCV\nfrom sklearn.metrics import classification_report\nfrom sklearn.metrics import accuracy_score, confusion_matrix\nfrom sklearn.model_selection import train_test_split, cross_val_score\n\n# + colab={\"base_uri\": \"https://localhost:8080/\", \"height\": 487} id=\"_H-PYh1OJco0\" outputId=\"b729e697-ab8c-4fdf-977b-694351b66c6f\"\n#import dataset\nforest=pd.read_csv('/content/forestfires.csv')\nforest\n\n# + [markdown] id=\"7sM_U0lFMBZD\"\n# EDA\n\n# + colab={\"base_uri\": \"https://localhost:8080/\", \"height\": 299} id=\"51utfiXiJr-l\" outputId=\"248c95f9-80db-4937-bf49-9f1c19dc4e3e\"\nforest.head()\n\n# + colab={\"base_uri\": \"https://localhost:8080/\", \"height\": 393} id=\"bdXZqzjoMHzw\" outputId=\"7dae874b-0d94-4f4c-e9cb-9226a72f033f\"\nforest.describe()\n\n# + colab={\"base_uri\": \"https://localhost:8080/\"} id=\"Fo_Lv4ceMQVi\" outputId=\"b6f6c48d-806d-4ee8-b5fc-e120bf3eaca6\"\nforest.info()\n\n# + colab={\"base_uri\": \"https://localhost:8080/\"} id=\"pMuQJNc9MWMs\" outputId=\"a1cf685a-b17c-4e0f-a626-77acdf04883c\"\nforest.isnull().sum()\n\n# + colab={\"base_uri\": \"https://localhost:8080/\", \"height\": 393} id=\"A_xxFnKfNTTj\" outputId=\"aa0105af-5929-4d80-c7c8-b7807ceff728\"\nforest[forest.duplicated()]\n\n# + colab={\"base_uri\": \"https://localhost:8080/\"} id=\"cBhuF2eZpACm\" outputId=\"6bb0b655-4421-4aeb-d9fd-46153d8eeb8c\"\nforest.pop('month') #pop is used to remove coloumns\nforest.pop('day')\n\n# + id=\"hw3P50GRONMU\"\nforest1=forest.drop_duplicates() #to deleting duplicate record\n\n# + colab={\"base_uri\": \"https://localhost:8080/\", \"height\": 487} id=\"eaCz_bs4OY5d\" outputId=\"ae0a0b12-59fd-4328-f6a7-6044c4cf4a7a\"\nforest1\n\n# + colab={\"base_uri\": \"https://localhost:8080/\"} id=\"uoSbciRITpQD\" outputId=\"49600ba2-54f0-422b-a19f-2a4ac82e3ce7\"\nforest1['size_category'].value_counts()\n\n# + colab={\"base_uri\": \"https://localhost:8080/\", \"height\": 471} id=\"b53ToeNJQr3q\" outputId=\"6c94d237-d45a-4642-9baa-fb93b9549e86\"\nforest1['size_category'].value_counts().plot.bar()\n\n# + id=\"x5mPTEQtRwFo\"\n#From above graph we can say most of forest fires where in small quantity \n\n# + colab={\"base_uri\": \"https://localhost:8080/\", \"height\": 1000} id=\"ntiSoibpcHO4\" outputId=\"b1b41add-5fbc-49aa-fade-aa7eb138c1eb\"\n# Drawing boxplot for indepent variables with continuous values\ncols = ['FFMC', 'DMC', 'DC', 'ISI', 'temp', 'RH', 'wind', 'area']\nplt.figure(figsize=(10,15))\n\ni=1\nfor col in cols:\n plt.subplot(4,2,i)\n sns.boxplot(y=col, data=forest,width=0.2)\n i+=1\nplt.show()\n\n# + colab={\"base_uri\": \"https://localhost:8080/\", \"height\": 484} id=\"WhMl3mdeZ4Wc\" outputId=\"ed891bd8-8ecb-4dc2-d7d3-5fb5ea231d17\"\npd.crosstab(forest['size_category'],forest['month']).mean().plot(kind='bar')\n\n# + id=\"BDjeUfcnaioS\"\n#above plot we conclude that majority of fire occurs in month August & Sep\n\n# + colab={\"base_uri\": \"https://localhost:8080/\", \"height\": 484} id=\"LaquWoDCdzbN\" outputId=\"2d3ffeee-2a6e-4efc-e1d7-3eb7bed23b97\"\npd.crosstab(forest['size_category'],forest['day']).mean().plot(kind='bar')\n\n# + id=\"LiSjFDJjfjUE\"\n# For Sunday & Friday have recored most cases of fire\n\n# + colab={\"base_uri\": \"https://localhost:8080/\", \"height\": 487} id=\"bSY86ntYiU-j\" outputId=\"fe679a72-97ea-4774-d7fc-4519401bda25\"\nforest1\n\n# + colab={\"base_uri\": \"https://localhost:8080/\", \"height\": 1000} id=\"s8JVSjNxgXOW\" outputId=\"274f0e6a-73ca-4601-8083-68bb6ae3deca\"\n#displaying pairplot to check relationship\nsns.pairplot(forest1,x_vars=['FFMC', 'DMC', 'DC', 'ISI', 'temp', 'RH', 'wind', 'area'],\n y_vars=['FFMC', 'DMC', 'DC', 'ISI', 'temp', 'RH', 'wind', 'area'],hue='size_category',\n kind='reg',diag_kind='kde')\nplt.show()\n\n# + colab={\"base_uri\": \"https://localhost:8080/\", \"height\": 449} id=\"4rlsI8xKjqxz\" outputId=\"5a9e9d81-10c9-4de4-ceee-cfa5e6bcd0cf\"\nsns.scatterplot(x='temp',y='RH',data=forest1,hue='size_category')\nplt.show()\n\n# + colab={\"base_uri\": \"https://localhost:8080/\", \"height\": 449} id=\"wmvPXQ1gkdXx\" outputId=\"7c93c046-2dc2-4603-e1e1-15f71c2349a3\"\nsns.scatterplot(x='temp',y='DMC',data=forest1,hue='size_category')\nplt.show()\n\n# + colab={\"base_uri\": \"https://localhost:8080/\", \"height\": 1000} id=\"tkm4NBQgk4ZY\" outputId=\"90ccd5da-7a27-4614-ebd6-f6d2fabcb650\"\ncor=forest.corr()\ncor\n\n# + colab={\"base_uri\": \"https://localhost:8080/\", \"height\": 897} id=\"HZMDCzYulpc1\" outputId=\"54f699a1-7ba6-4614-da2e-1425c579c512\"\n#checking corelation \nplt.figure(figsize=(12,10))\nsns.heatmap(cor, annot=True,)\nplt.show()\n\n# + [markdown] id=\"12WbYUnum40t\"\n# Feature Engineering\n\n# + id=\"tvQx68btmPYu\"\narray = forest1.values\nx = forest1.iloc[:,0:28]\ny = forest1.iloc[:,28]\n\n# + colab={\"base_uri\": \"https://localhost:8080/\", \"height\": 487} id=\"M9abwFzVp3PB\" outputId=\"9edaf0fa-face-4839-a0e1-88f2d51f1979\"\nx\n\n# + colab={\"base_uri\": \"https://localhost:8080/\"} id=\"qS9evjnfqtcV\" outputId=\"5c425623-c372-4111-ae3b-3e0227fb46d4\"\ny\n\n\n# + id=\"U3lpaaRFq89C\"\n#Normalization\ndef norm_func(i):\n x = (i-i.min())/(i.max()-i.min())\n return (x)\n\n\n# + id=\"oAVYNCsGrZDM\"\nforestfires=norm_func(x)\n\n# + colab={\"base_uri\": \"https://localhost:8080/\", \"height\": 487} id=\"F6NtOfpzrlBO\" outputId=\"1fe47d07-4e90-4863-aee6-d123b8279f8e\"\nforestfires\n\n# + id=\"iJvXUYAKr0Db\"\nX_train, X_test, y_train, y_test = train_test_split(x,y, test_size = 0.3)\n\n# + colab={\"base_uri\": \"https://localhost:8080/\"} id=\"22SZ_1per95Y\" outputId=\"71bac0a8-f7f3-449d-84d4-de85f8397378\"\nX_train.shape, y_train.shape, X_test.shape, y_test.shape\n\n# + [markdown] id=\"k_45YPKXsgVW\"\n# SVM Model\n\n# + colab={\"base_uri\": \"https://localhost:8080/\", \"height\": 74} id=\"Z9UBJ-kksjvr\" outputId=\"9d94a9af-ad09-4fa6-e1b1-cf5ced796ec0\"\n#Kernel=Linear\nmodel_linear = SVC(kernel = \"linear\")\nmodel_linear.fit(X_train,y_train)\n\n# + id=\"HEL7fEdGtA50\"\npred_test_linear = model_linear.predict(X_test)\n\n# + colab={\"base_uri\": \"https://localhost:8080/\"} id=\"SCsnhd36tJBS\" outputId=\"dc65190a-fa3a-4982-b90d-b2c71e9369ec\"\nnp.mean(pred_test_linear==y_test) #97.38% accuracy\n\n# + colab={\"base_uri\": \"https://localhost:8080/\"} id=\"EIAs9DHntmMK\" outputId=\"6e5a5be1-b1ea-498f-9413-269ce183b7f3\"\nacc = accuracy_score(y_test, pred_test_linear) * 100\nprint(\"Accuracy =\", acc)\nconfusion_matrix(y_test, pred_test_linear)\n\n# + colab={\"base_uri\": \"https://localhost:8080/\", \"height\": 74} id=\"uyO0_1CLtsos\" outputId=\"b5c4d0e1-eb34-408a-b315-7cf1fd647bd0\"\n#kernel=Poly\nmodel_poly = SVC(kernel = \"poly\")\nmodel_poly.fit(X_train,y_train)\n\n# + id=\"5b_dKcHet-WZ\"\npred_test_poly = model_poly.predict(X_test)\n\n# + colab={\"base_uri\": \"https://localhost:8080/\"} id=\"9n2G6d2XuK51\" outputId=\"eda40ff3-6ef4-473e-cf09-6674c2e126e2\"\nnp.mean(pred_test_poly==y_test) #73.85% accuracy\n\n# + colab={\"base_uri\": \"https://localhost:8080/\"} id=\"L3O9IOGduc_L\" outputId=\"025e99d8-77b6-4dab-9eb8-d88fdcb4c8e4\"\nacc = accuracy_score(y_test, pred_test_linear) * 100\nprint(\"Accuracy =\", acc)\nconfusion_matrix(y_test, pred_test_linear)\n\n# + colab={\"base_uri\": \"https://localhost:8080/\", \"height\": 74} id=\"CaHhXH5Huw6i\" outputId=\"7380f360-5beb-4b0c-d726-d3c21b2141a0\"\n#kernel=rbf\nmodel_rbf = SVC(kernel = \"rbf\")\nmodel_rbf.fit(X_train,y_train)\n\n# + id=\"s0bIBB4wvGEy\"\npred_test_rbf = model_rbf.predict(X_test)\n\n# + colab={\"base_uri\": \"https://localhost:8080/\"} id=\"HQtJhkXovNlF\" outputId=\"4adc47f4-b9fa-4b91-8ceb-0fe833f7e880\"\nnp.mean(pred_test_rbf==y_test) #Accuracy =72.54%\n\n\n# + colab={\"base_uri\": \"https://localhost:8080/\", \"height\": 74} id=\"l06u-quRvf0m\" outputId=\"43979ce3-4c3d-40aa-bb32-d98a53afe4f3\"\n#Kernel=Sigmoid\nmodel_sig = SVC(kernel = \"sigmoid\")\nmodel_sig.fit(X_train,y_train)\n\n# + id=\"jKT0O5Glvk5X\"\npred_test_sig = model_sig.predict(X_test)\n\n# + colab={\"base_uri\": \"https://localhost:8080/\"} id=\"WBrYbYAqvuB1\" outputId=\"03cca709-a5a8-4fc3-a1f8-7e833b520c89\"\nnp.mean(pred_test_sig==y_test) \n","repo_name":"aschobhe/data-science-assignments","sub_path":"Asssignment_17_SVM_Forestfires.ipynb","file_name":"Asssignment_17_SVM_Forestfires.ipynb","file_ext":"py","file_size_in_byte":7925,"program_lang":"python","lang":"en","doc_type":"code","stars":0,"dataset":"github-jupyter-script","pt":"44"}
+{"seq_id":"35981427583","text":"# +\n# The imports we need for this project are as follows:\n# import libraries needed\nimport pandas as pd\nimport numpy as np\nimport matplotlib.pyplot as plt\nimport seaborn as sns\n\nimport spotipy\nfrom spotipy.oauth2 import SpotifyClientCredentials\n\nimport warnings\nwarnings.filterwarnings(\"ignore\")\n# -\n\ncid = \"XXX\"\nsecret = \"XXX\"\n#Authentication - without user\nclient_credentials_manager = SpotifyClientCredentials(client_id=cid, client_secret=secret)\nsp = spotipy.Spotify(client_credentials_manager = client_credentials_manager)\n\ndf = pd.read_csv('top_200_weekly.csv')\ndf.head()\n\ndf.nunique()\n\n# ### Missing value\n\ndf.info()\n\ndf.loc[df['Artist'].isna()].style.highlight_null(null_color='darkgreen')\n\n# +\n# Toy exxample get \n# track = 'spotify:track:4iJyoBOLtHqaGxP12qzhQI'\n# track = sp.track(track)\n# # print(track.keys())\n# print(track['name'])\n# print([a['name'] for a in track['artists']])\n\ntracks = list(df.loc[df['Artist'].isna()]['URI'])\nfor track in tracks:\n print('*'*20)\n print(track)\n track = sp.track(track)\n\n print(track['name'],[a['name'] for a in track['artists']]) \n# -\n\ndf.nunique()\n\n# ### New datasets\n\ndf_rank = df[['Week End','ID','Position']]\n# df_rank = df_rank.sort_values(by=['Week End','Position'])\ndf_rank.head()\n\ndf_rank_pivot = df_rank.pivot(index=['Week End'],columns=['Position'],values=['ID'])\ndf_rank_pivot.to_csv('week_position.csv',index_label=False)\n\ndf_rank_pivot.head()\n\n# ### Q1. What is the pattern of longevity of success songs?\n\ndf_1 = df.copy()\n\ndf_1.nunique()\n\ndf_1[['ID','URI']].groupby(by=['ID']).count()['URI'].sort_values()[-5:]\n\ndf_1[df_1.ID=='7qiZfU4dY1lWllzX7mPBI3'].head(1)\n\ndf_1[df_1.ID=='5uCax9HTNlzGybIStD3vDh'].head(1)\n\ndf_1[df_1.ID=='0tgVpDi06FyKpA1z0VMD4v'].head(1)\n\ndf_1[df_1.ID=='6gBFPUFcJLzWGx4lenP6h2'].head(1)\n\ndf_1[df_1.ID=='7m9OqQk4RVRkw9JJdeAw96'].head(1)\n\ndata = list(df_1[['ID','URI']].groupby(by=['ID']).count()['URI'])\nprint(data[:10])\nprint(f\"Number of tracks: {len(data)}\")\nprint(f\"\")\n\ndf_1[['ID','URI']].groupby(by=['ID']).count()\n\n# +\n# plt.figure(figsize=(10,5))\n# plt.title('Longevity of bestsellers',loc='center')\nfig, ax = plt.subplots(figsize=(7,5))\n# hist, bins, _ = plt.hist(data, 50, facecolor='g', alpha=0.75)\nplt.hist(data, 50, facecolor='g',edgecolor=\"black\", alpha=0.7)#,rwidth=0.85)\n# top1\nax.annotate(text='', xy=(257, 1), xycoords='data',\n xytext=(-50, 230), textcoords='offset points',#'axes fraction',\n arrowprops=dict(arrowstyle='->',facecolor='black'),#, shrink=0.05,width=0.3,headwidth=3),\n # horizontalalignment='right', verticalalignment='top',\n # textprops=dict()\n )\nax.text(180, 2700, 'Shape of You \\nEd Sheeran \\n257 weeks', style='italic',fontweight='medium')\n # bbox={'facecolor': 'g', 'alpha': 0.5, 'pad': 10})\n# top2\nax.annotate(text='', xy=(255, 1), xycoords='data',\n xytext=(-80, 175), textcoords='offset points',#'axes fraction',\n arrowprops=dict(arrowstyle='->',facecolor='black'),#, shrink=0.05,width=0.3,headwidth=3),\n )\nax.text(140, 2000, \"Say You Won't Let Go \\nJames Arthur \\n255 weeks\", style='italic',fontweight='medium')\n # bbox={'facecolor': 'g', 'alpha': 0.35, 'pad': 10})\n\n# top3\nax.annotate(text='', xy=(250, 1), xycoords='data',\n xytext=(-100, 110), textcoords='offset points',#'axes fraction',\n arrowprops=dict(arrowstyle='->',facecolor='black'),#, shrink=0.05,width=0.3,headwidth=3),\n )\nax.text(150, 1400, \"Perfect \\nEd Sheeran \\n250 weeks\", style='italic',fontweight='medium')\n # bbox={'facecolor': 'g', 'alpha': 0.2, 'pad': 10})\n\n# top4\nax.annotate(text='', xy=(224, 1), xycoords='data',\n xytext=(-120, 110), textcoords='offset points',#'axes fraction',\n arrowprops=dict(arrowstyle='->',facecolor='black'),#, shrink=0.05,width=0.3,headwidth=3),\n )\nax.text(90, 1200, \"Goosebumps \\nTravis Scott \\n224 weeks\", style='italic',fontweight='medium')\n\n# top5\nax.annotate(text='', xy=(215, 1), xycoords='data',\n xytext=(-150, 70), textcoords='offset points',#'axes fraction',\n arrowprops=dict(arrowstyle='->',facecolor='black'),#, shrink=0.05,width=0.3,headwidth=3),\n )\nax.text(50, 700, \"Jocelyn Flores \\nXXXTENTACION \\n215 weeks\", style='italic',fontweight='medium')\n \n# ID\n# 7m9OqQk4RVRkw9JJdeAw96 215\n# 6gBFPUFcJLzWGx4lenP6h2 224\n# 0tgVpDi06FyKpA1z0VMD4v 250\n# 5uCax9HTNlzGybIStD3vDh 255\n# 7qiZfU4dY1lWllzX7mPBI3 257\nplt.xlabel('Number of Weeks on Charts',fontsize=11,fontweight='medium')\nplt.ylabel('Number of Tracks',fontsize=11,fontweight='medium') \n# histogram on log scale. \n# Use non-equal bin sizes, such that they look equal on log scale.\n# logbins = np.logspace(np.log10(bins[0]),np.log10(bins[-1]),len(bins))\n# plt.title('Longevity of Success Tracks',loc='center')\nplt.savefig('figures/figure0.png',dpi=300)\n\nplt.show()\n# -\n\nfig, ax = plt.subplots(figsize=(7,5))\nplt.hist(data, bins=50,facecolor='b',edgecolor=\"black\", alpha=0.7,log=True)#,histtype=\"step\",cumulative=True)\n# plt.hist(data, alpha=0.5,log=True,histtype=\"step\",cumulative=True)\n# plt.xscale('log')\n# plt.yscale('log')\nplt.xlabel('Number of Weeks on Charts',fontsize=11,fontweight='medium')\nplt.ylabel('Number of Tracks (log)',fontsize=11,fontweight='medium')\n# plt.title('Longevity of Success Tracks (log)',loc='center')\n# plt.tight_layout()\nax.text(210, 1.1, \"1\", style='italic',fontweight='bold',c='black',fontsize=12)\nax.text(220, 1.1, \"1\", style='italic',fontweight='bold',c='black',fontsize=12)\nax.text(245, 1.1, \"1\", style='italic',fontweight='bold',c='black',fontsize=12)\nax.text(252, 2.1, \"2\", style='italic',fontweight='bold',c='black',fontsize=12)\nplt.savefig('figures/figure1.png',dpi=300)\nplt.show()\n\nnumber_of_week = list(df_1.groupby(by=['ID'])['Position'].count())\nbest_rank = list(df_1.groupby(by=['ID'])['Position'].min())\n\ndf_1.groupby(by=['ID'])['Position'].count()\n\ndf_1.groupby(by=['ID'])['Position'].min()\n\n# +\n# plt.figure(figsize=(7,5))\n# plt.scatter(number_of_week,best_rank,c='red',alpha=0.7,linewidths=0.0000001)\n# plt.ylim((201, 0))\n# # plt.ylim((0, 201))\n# plt.xscale('log')\n# # plt.yscale('log')\n# # plt.title('Best rank vs Number of weeks stayes')\n# plt.xlabel('Number of weeks on Global Top 200')\n# plt.ylabel('Best rank')\n# plt.tight_layout()\n\n# plt.savefig('figures/figure2.png')\n# plt.show()\n\n# +\nfrom collections import defaultdict\n\ncombination = list(zip(number_of_week,best_rank))\n\ncounter = defaultdict(int)\n\nfor c in combination:\n # x.append(c[0])\n # y.append(c[1])\n counter[c] += 1\n \n\nx = [i[0] for i in counter.keys()]\ny = [i[1] for i in counter.keys()]\ns = list(counter.values())\nsize = [np.pi*i*i for i in s]\nlen(size),len(x),len(y)\n\n# +\nplt.figure(figsize=(7,5))\nplt.scatter(x,y,s=size,c=\"blue\",alpha=0.7)\nplt.ylim((201, 0))\n# plt.ylim((0, 201))\nplt.xscale('log')\n\n# plt.title('Best rank vs Number of weeks stayes')\nplt.xlabel('Number of Weeks on Charts',fontsize=11,fontweight='medium')\nplt.ylabel('Best Rank',fontsize=11,fontweight='medium')\n# plt.tight_layout()\n\nplt.savefig('figures/figure2.png',dpi=300)\nplt.show()\n\n# +\n# GO DEEPER OF THE TOP FIVE\n# 7m9OqQk4RVRkw9JJdeAw96 215\n# 6gBFPUFcJLzWGx4lenP6h2 224\n# 0tgVpDi06FyKpA1z0VMD4v 250\n# 5uCax9HTNlzGybIStD3vDh 255\n# 7qiZfU4dY1lWllzX7mPBI3 257\n# -\n\ndf = pd.read_csv('top_200_weekly.csv')\ndf.head()\n\ndf_top5 = df[(df['ID']=='7qiZfU4dY1lWllzX7mPBI3') | \n (df['ID']=='5uCax9HTNlzGybIStD3vDh') |\n (df['ID']=='0tgVpDi06FyKpA1z0VMD4v') |\n (df['ID']=='6gBFPUFcJLzWGx4lenP6h2') |\n (df['ID']=='7m9OqQk4RVRkw9JJdeAw96')\n ]\ndf_top5.head()\n\ndf_top5.loc[:,'Year'] = pd.Series(map(lambda x:int(x[:4]) ,df['Week End']))\ndf_top5.loc[:,'Month'] = pd.Series(map(lambda x:int(x[5:7]) ,df['Week End']))\n\ndf_top5.info()\ndf_top5.head()\n\ndf_top5.to_csv('top5.csv', index=False)\n\ndf_top5 = pd.read_csv('top5.csv',parse_dates=['Week End'])\n# df_top5.set_index('Week End', inplace=True)\ndf_top5.head()\n\nids = ['7m9OqQk4RVRkw9JJdeAw96',\n'6gBFPUFcJLzWGx4lenP6h2',\n'0tgVpDi06FyKpA1z0VMD4v',\n'5uCax9HTNlzGybIStD3vDh',\n'7qiZfU4dY1lWllzX7mPBI3']\nprint(f\"isd:{ids}\")\n\nfor i,t in enumerate(ids[::-1]):\n _1 = df_top5[df_top5['ID'] == t]\n df_1 = _1.groupby(by=['Year','Month']).min().reset_index()\n position_1 = df_1.pivot( \"Year\",\"Month\", \"Position\").fillna(0).astype('int64')\n # heatmap\n fig, ax = plt.subplots(figsize=(8,3))\n mask = position_1==0\n sns.heatmap(position_1, cmap=\"summer\",annot=True, annot_kws={\"size\": 10},fmt='d',\n linewidths=0.5, linecolor='white',ax=ax,mask=mask,cbar_kws={'label':'Best Position'})\n plt.tight_layout()\n path = f\"figures/top5_{i+1}.png\"\n plt.savefig(path,dpi=300)\n plt.show()\n # break\n\n_1\n\nsns.set_style(\"white\")\nplt.figure(figsize=(12,6))\nplt.ylim((201, 0))\npalette = [\"green\",\"orange\",\"red\",\"blue\",\"purple\"]\nsns.lineplot(x='Week End',y='Position',data=df_top5,hue='Track Name',palette=palette,marker='o')\nplt.xlabel('Date',size=12)\nplt.tight_layout()\nplt.savefig(\"figures/top5.png\",dpi=300)\nplt.show()\n\n\n\n# ### Q6. What are the relations between position, length of stay of a song on the list and song streams? \n\ndf = pd.read_csv('top_200_weekly.csv')\ndf.head()\n\ndata_streams = df.groupby(by='ID').max()['Streams']\n\n# +\nfig, ax = plt.subplots(figsize=(7,5))\n# plt.hist(data_streams,bins=100,facecolor='g', alpha=0.7)\nplt.hist(data_streams, 100, facecolor='green',edgecolor=\"black\", alpha=0.7)#,rwidth=0.85)\nplt.xscale('log')\n# plt.yscale('log')\n# plt.title('Number of songs vs Streams')\nplt.xlabel('Streams')\nplt.ylabel('Number od songs')\nplt.tight_layout()\n\nplt.savefig('figures/figure3.png',dpi=300)\nplt.show()\nplt.show()\n\n# +\nfig, ax = plt.subplots(figsize=(7,5))\nplt.scatter(best_rank,data_streams,facecolor='blue', alpha=0.7)#,c='DarkBlue')\n# plt.ylim((201, 0))\n# plt.xscale('log')\n# plt.yscale('log')\n# plt.title('Best rank vs Number of weeks stayes')\nplt.xlabel('Best rank')\nplt.ylabel('Streams')\nplt.tight_layout()\n\nplt.savefig('figures/figure4.png',dpi=300)\nplt.show()\n\n# +\nfig, ax = plt.subplots(figsize=(7,5))\nplt.scatter(number_of_week,data_streams,facecolor='red', alpha=0.7)#,c='DarkBlue')\n# sns.scatterplot(number_of_week,data_streams,data=df_1)\n# plt.ylim((201, 0))\n# plt.xscale('log')\nplt.yscale('log')\n# plt.title('Best rank vs Number of weeks stayes')\nplt.xlabel('Number of weeks stayes')\nplt.ylabel('Streams')\nplt.tight_layout()\n\nplt.savefig('figures/figure5.png',dpi=300)\nplt.show()\n\n# +\n# from random import randint\n# # %matplotlib tk\n# import matplotlib.pyplot as plt\n# from matplotlib.animation import FuncAnimation\n# from IPython.display import HTML\n# # create empty lists for the x and y data\n# x = []\n# y = []\n\n# # create the figure and axes objects\n# fig, ax = plt.subplots()\n\n# # function that draws each frame of the animation\n# def animate(i):\n# pt = randint(1,9) # grab a random integer to be the next y-value in the animation\n# x.append(i)\n# y.append(pt)\n\n# ax.clear()\n# ax.plot(x, y)\n# ax.set_xlim([0,20])\n# ax.set_ylim([0,10])\n \n \n# # run the animation\n# ani = FuncAnimation(fig, animate, frames=20, interval=500, repeat=False)\n# # HTML(ani.to_html5_video())\n# ani.save(\"sin1.gif\",writer='pillow')\n# plt.show()\n","repo_name":"chennnxu/HitsOnSpotify","sub_path":"3_analysis_tracks.ipynb","file_name":"3_analysis_tracks.ipynb","file_ext":"py","file_size_in_byte":11205,"program_lang":"python","lang":"en","doc_type":"code","stars":0,"dataset":"github-jupyter-script","pt":"44"}
+{"seq_id":"73443458692","text":"# + colab={} colab_type=\"code\" id=\"RQr9JaI_mkfo\"\nfrom sklearn.linear_model import LogisticRegression\nfrom sklearn.metrics import accuracy_score\nimport matplotlib.pyplot as plt\nimport pandas as pd\n\n# + colab={} colab_type=\"code\" id=\"wN1o9zjrr8HQ\"\ndf = pd.read_csv('diabetes.csv')\n\n# + colab={} colab_type=\"code\" id=\"EPdby81lEVcf\"\nX = df.iloc[:, 0:7]\ny = df['Outcome']\n\nfrom sklearn.model_selection import train_test_split\nX_train, y_train, X_test, y_test = train_test_split(X, y, test_size = 0.20)\n\n# + colab={} colab_type=\"code\" id=\"GRzf_CTDEVUz\"\nmodel = LogisticRegression()\n\n# + colab={} colab_type=\"code\" id=\"fUT2xp_eEVQl\"\nmodel.fit(X_train, y_train)\n\n# + colab={} colab_type=\"code\" id=\"rJm-LCgtEVK-\"\n# coefficients of the trained model\nprint('Coefficient of model :', model.coef_)\n\n# intercept of the model\nprint('Intercept of model',model.intercept_)\n\n# + colab={} colab_type=\"code\" id=\"PBACA9g7E2rD\"\n# predict the target on the train dataset\npredict_train = model.predict(train_x)\nprint('Target on train data',predict_train) \n\n# Accuray Score on train dataset\naccuracy_train = accuracy_score(train_y,predict_train)\nprint('accuracy_score on train dataset : ', accuracy_train)\n\n# predict the target on the test dataset\npredict_test = model.predict(test_x)\nprint('Target on test data',predict_test) \n\n# Accuracy Score on test dataset\naccuracy_test = accuracy_score(test_y,predict_test)\nprint('accuracy_score on test dataset : ', accuracy_test)\n\n# + [markdown] colab_type=\"text\" id=\"bfCmvqJhsg-X\"\n# ##Fitting a model between two columns col1 and col2:\n# col1 contains the basis of classification and\n# col2 contains the classification result(1 or 0)\n\n# + colab={} colab_type=\"code\" id=\"2ajdXzItr8Qr\"\nlogistic_model = LogisticRegression()\nlogistic_model.fit(dataframe[[\"col1\"]], dataframe[\"col2\"])\nfitted_labels = logistic_model.predict(dataframe[['col1']])\nplt.scatter(dataframe['col1'], fitted_labels)\n","repo_name":"sarthakrastogi/notesonnotebooks","sub_path":"ML Algorithms/Logistic_Regression.ipynb","file_name":"Logistic_Regression.ipynb","file_ext":"py","file_size_in_byte":1905,"program_lang":"python","lang":"en","doc_type":"code","stars":0,"dataset":"github-jupyter-script","pt":"44"}
+{"seq_id":"72762459320","text":"# + id=\"zlU5dsSk3-HT\"\nimport pandas as pd\nimport seaborn as sns\nimport statsmodels.api as sm\nimport numpy as np\nfrom scipy.stats import pearsonr\nfrom scipy import stats\n\n# + id=\"caUYF-oy3_nb\"\nboston_url = 'https://cf-courses-data.s3.us.cloud-object-storage.appdomain.cloud/IBMDeveloperSkillsNetwork-ST0151EN-SkillsNetwork/labs/boston_housing.csv'\nboston_df=pd.read_csv(boston_url)\n\n# + colab={\"base_uri\": \"https://localhost:8080/\", \"height\": 206} id=\"zKMt-MVo4AB5\" outputId=\"ef9d7d01-a3ef-4cc4-e4cc-7453782e8b7b\"\nboston_df.head()\n\n# + colab={\"base_uri\": \"https://localhost:8080/\", \"height\": 350} id=\"p29BI56Y694D\" outputId=\"4945d3ed-659d-42d1-b1c2-275d23e3dc40\"\n#For the \"Median value of owner-occupied homes\" provide a boxplot\nimport matplotlib.pyplot as plt\n# %matplotlib inline\nplt.figure(figsize=(10,5))\nsns.boxplot(x=boston_df.MEDV)\nplt.title(\"Boxplot for MEDV\")\nplt.show()\n\n# + colab={\"base_uri\": \"https://localhost:8080/\", \"height\": 405} id=\"D05XIXS44OLh\" outputId=\"1d7f47f6-73a7-4d38-d7b4-c7c5b46b3ce4\"\nplt.figure(figsize=(10,5))\nsns.distplot(a=boston_df.CHAS,bins=10, kde=False)\nplt.title(\"Histogram for Charles river\")\nplt.show()\n\n# + colab={\"base_uri\": \"https://localhost:8080/\", \"height\": 350} id=\"ZlUs7Dc64glU\" outputId=\"429cf6f3-f6fa-4f65-bd0c-ef8cdfb8fd09\"\nplt.figure(figsize=(10,5))\nsns.histplot(data=boston_df, x=\"PTRATIO\")\nplt.title(\"Histogram Plot for the PT ratio\")\nplt.show()\n\n# + colab={\"base_uri\": \"https://localhost:8080/\", \"height\": 350} id=\"eskw23s-5aSe\" outputId=\"358e8dbb-4793-44d4-e636-8b8a336bac59\"\nplt.figure(figsize=(10,5))\nsns.scatterplot(data=boston_df, x=\"NOX\", y=\"INDUS\")\nplt.title(\"ScatterPlot for the NOX variable vs the INDUS variable\")\nplt.show()\n\n# + colab={\"base_uri\": \"https://localhost:8080/\"} id=\"NUib3G-I5qXu\" outputId=\"741a276c-8f5b-4b9e-bf7e-b2dfb81e6178\"\nlist1 = boston_df['NOX']\nlist2 = boston_df['INDUS'] \n# Apply the pearsonr()\ncorr, _ = pearsonr(list1, list2)\nprint('Pearsons correlation: %.3f' % corr)\nprint('From the above firgure, it can be seen that NOX and INDUS are correlated')\n\n# + id=\"Bs5c7t-26NU9\"\nboston_df.loc[(boston_df[\"AGE\"] <= 35),'age_group'] = '35 years and younger'\nboston_df.loc[(boston_df[\"AGE\"] > 35) & (boston_df[\"AGE\"]<70),'age_group'] = 'between 35 and 70 years'\nboston_df.loc[(boston_df[\"AGE\"] >= 70),'age_group'] = '70 years and older'\n\n# + colab={\"base_uri\": \"https://localhost:8080/\", \"height\": 350} id=\"UMd5knmf9Kwr\" outputId=\"c36024a8-ca69-4ac5-aad3-9667c52a05c8\"\nplt.figure(figsize=(10,5))\nsns.boxplot(x=boston_df.MEDV, y=boston_df.age_group, data=boston_df)\nplt.title(\"Boxplot for the MEDV variable vs the AGE variable\")\nplt.show()\n\n# + [markdown] id=\"0srF0TEB99R2\"\n# For each of the following questions;\n#\n# Is there a significant difference in median value of houses bounded by the Charles river or not? (T-test for independent samples)\n#\n# Is there a difference in Median values of houses (MEDV) for each proportion of owner occupied units built prior to 1940 (AGE)? (ANOVA)\n#\n# Can we conclude that there is no relationship between Nitric oxide concentrations and proportion of non-retail business acres per town? (Pearson Correlation)\n#\n# What is the impact of an additional weighted distance to the five Boston employment centres on the median value of owner occupied homes? (Regression analysis)\n#\n# Be sure to:\n#\n# State your hypothesis.\n#\n# Use α = 0.05\n#\n# Perform the test Statistics.\n#\n# State the conclusion from the test.\n\n# + [markdown] id=\"w0lyl_tB_7jB\"\n# **Is there a significant difference in median value of houses bounded by the Charles river or not? (T-test for independent samples)**\n#\n#\n\n# + colab={\"base_uri\": \"https://localhost:8080/\"} id=\"2RONDtwa-T2G\" outputId=\"14a02653-6979-4d94-844d-8ac25e9c61be\"\nrvs1=boston_df.loc[(boston_df['CHAS'] ==1),'MEDV']\nrvs2=boston_df.loc[(boston_df['CHAS'] ==0),'MEDV']\nstats.ttest_ind(rvs1, rvs2)\n\n# + [markdown] id=\"UBNJHdUyAACA\"\n# Null Hypothesis(): Both average MEDV are the same\n#\n# Alternative Hypothesis(): Both average MEDV are NOT the same\n#\n# Since alpha =0.05 and the pvalue < alpha. Therefore, the we reject NULL hypothesis\n#\n\n# + [markdown] id=\"q_a8Yj65ArYQ\"\n#\n# **Is there a difference in Median values of houses (MEDV) for each proportion of owner occupied units built prior to 1940 (AGE)? (ANOVA)**\n\n# + colab={\"base_uri\": \"https://localhost:8080/\"} id=\"_hZBxlfx_K1z\" outputId=\"1e3eda9c-2cca-43ad-8d59-08c43830903f\"\nlow = boston_df[boston_df[\"age_group\"] == '35 years and younger'][\"MEDV\"]\nmid = boston_df[boston_df[\"age_group\"] == 'between 35 and 70 years'][\"MEDV\"]\nhigh = boston_df[boston_df[\"age_group\"] == '70 years and older'][\"MEDV\"]\nstats.f_oneway(low, mid, high)\n\n# + [markdown] id=\"gaZanHORLMvx\"\n# Th null hypothesis is the hypothesis that all age groups have same mean MEDV values. Because the pvalue < alpha, therefore, the NULL hypothesis is rejected.\n\n# + [markdown] id=\"1wfCy5viLrJb\"\n# **Can we conclude that there is no relationship between Nitric oxide concentrations and proportion of non-retail business acres per town? (Pearson Correlation)**\n\n# + colab={\"base_uri\": \"https://localhost:8080/\"} id=\"r6vDhCimKjqc\" outputId=\"6ea6e3b8-3892-4854-d924-394cbad5dc46\"\nNOX_data=boston_df['NOX']\nINDUS_data=boston_df['INDUS']\nres = stats.pearsonr(NOX_data,INDUS_data )\nres\n\n# + [markdown] id=\"GztdrTpzLqSS\"\n# The Null hypothesis is that there is no relationship between NOX and INDUS variable. However, the pvalue < alpha. Therefore, the NULL hypothesis is rejected. This can also be conlcuded from scateerplot plot\n\n# + [markdown] id=\"8hMBnbz9MkFV\"\n# **What is the impact of an additional weighted distance to the five Boston employment centres on the median value of owner occupied homes? (Regression analysis)**\n\n# + colab={\"base_uri\": \"https://localhost:8080/\", \"height\": 498} id=\"HyGNzFEXMhZU\" outputId=\"3f94dfc9-a167-4ed0-def1-2dd24b14adec\"\nx=boston_df['DIS']\ny=boston_df['MEDV']\nx = sm.add_constant(x)\nmodel = sm.OLS(y,x)\nresults = model.fit()\nresults.summary()\n\n# + [markdown] id=\"ipGo9MZCNmRD\"\n# **The DIS parameter add 1.0916 units more than average to MEDV**\n\n# + id=\"Pbu3bLeENfHs\"\n\n","repo_name":"sreenivasraghavan71/Coursera","sub_path":"Assignment.ipynb","file_name":"Assignment.ipynb","file_ext":"py","file_size_in_byte":6068,"program_lang":"python","lang":"en","doc_type":"code","stars":0,"dataset":"github-jupyter-script","pt":"40"}
+{"seq_id":"32200546479","text":"# %matplotlib notebook\n\nfrom create_ice_cube_image_data import *\nfrom numba import jit\nfbatch={'1':\"/scratch/ICE-CUBE/batch_1.parquet\"}\nice_cube=ice_cube_data(fbatch=fbatch,\n fsensor=\"/scratch/ICE-CUBE/sensor_geometry.csv\",\n fmetaf=\"/scratch/ICE-CUBE/train_meta.parquet\",\n every_nth_event=10,\n width={'x':10,'y':10,'z':60})\n\n# +\n#import matplotlib.pyplot as plt\n#threedee = plt.figure(figsize=(5, 5)).gca(projection='3d')\n#threedee.scatter(ice_cube.events['x'],ice_cube.events['y'],ice_cube.events['z'])\n#plt.show()\n# -\n\nimg_data, labels= ice_cube.create_img_data()\nimg_data.shape\n\neventdict={}\nfor ievent,indexevent in enumerate(ice_cube.events.index.unique().to_list()):\n eventdict[ievent]=indexevent\n \n\neventdraw=200\n\n# +\nimport matplotlib.pyplot as plt\nfig = plt.figure()\nax = fig.add_subplot(111, projection='3d')\n\n# create a meshgrid for the x, y, and z coordinates\n#x, y, z = np.meshgrid(np.arange(img_width), np.arange(img_height), np.arange(img_depth))\nx,y,z,=img_data[eventdraw].nonzero()\n\n# plot the 3D image using the meshgrid and the pixel values\nax.scatter(x,y,z)#, c=x_train[0].flatten(),cmap='gray_r',s=20)\n\n# set the axis labels and limits\nax.set_xlabel('X')\nax.set_ylabel('Y')\nax.set_zlabel('Z')\nax.set_xlim([0, 10])\nax.set_ylim([0, 10])\nax.set_zlim([0, 60])\n\n# show the 3D plot\nplt.show()\n\n# +\nimport matplotlib.pyplot as plt\nfig = plt.figure()\nax = fig.add_subplot(111, projection='3d')\n\n# plot the 3D image using the meshgrid and the pixel values\nax.scatter(ice_cube.events.loc[eventdict[eventdraw]]['x'],ice_cube.events.loc[eventdict[eventdraw]]['y'],ice_cube.events.loc[eventdict[eventdraw]]['z'])\n\n# set the axis labels and limits\nax.set_xlabel('X')\nax.set_ylabel('Y')\nax.set_zlabel('Z')\nax.set_xlim([0, 10])\nax.set_ylim([0, 10])\nax.set_zlim([0, 60])\n\n# show the 3D plot\nplt.show()\n\n# +\nimport matplotlib.pyplot as plt\nfig = plt.figure()\nax = fig.add_subplot(111, projection='3d')\n\n# plot the 3D image using the meshgrid and the pixel values\nax.scatter(ice_cube.realevents.loc[eventdict[eventdraw]]['x'],ice_cube.realevents.loc[eventdict[eventdraw]]['y'],ice_cube.realevents.loc[eventdict[eventdraw]]['z'])\n\n# set the axis labels and limits\nax.set_xlabel('X')\nax.set_ylabel('Y')\nax.set_zlabel('Z')\nax.set_xlim([-500, 500])\nax.set_ylim([-500, 500])\nax.set_zlim([-500, 500])\n\n# show the 3D plot\nplt.show()\n# -\n\nimg_data=img_data.reshape(19976, 11, 11, 61,1)\nimg_data.shape\n\nfrom sklearn.model_selection import train_test_split\nX_train, X_test, y_train, y_test = train_test_split(img_data, labels, test_size=0.9, random_state=42)\n\n# +\nimport tensorflow as tf\nimport os\nos.environ[\"CUDA_VISIBLE_DEVICES\"] = \"-1\"\n\nfrom tensorflow.keras.models import Sequential\nfrom tensorflow.keras.layers import Conv3D, MaxPooling3D, Flatten, Dense\n\n# create a sequential model\nmodel = Sequential()\n\n# add a 3D convolutional layer with 32 filters, a 3x3x3 kernel size, and ReLU activation\nmodel.add(Conv3D(32, (3, 3, 3), activation='relu', input_shape=X_train[0].shape))\n\n# add a 3D max pooling layer with a 2x2x2 pool size\nmodel.add(MaxPooling3D(pool_size=(2, 2, 2)))\n\n# flatten the output of the previous layer\nmodel.add(Flatten())\n\n# add a fully connected (Dense) layer with 128 units and ReLU activation\nmodel.add(Dense(100, activation='relu'))\n\n# add the output layer with 2 units (assuming 2 classes) and softmax activation\nmodel.add(Dense(2, activation='relu'))\n\n# compile the model with categorical crossentropy loss and Adam optimizer\nmodel.compile(loss='mse', optimizer='adam',metrics=[\"categorical_accuracy\"])\n\n# assuming X_train contains the image data and y_train contains the labels, fit the model\nmodel.fit(X_train, y_train, epochs=10, batch_size=32)\n# -\n\nmodel.predict(X_test)\n\ny_test\n\nmodel.predict(X_train)\n\ny_train\n","repo_name":"Exploratory-Machine-Learning/ICE-CUBE","sub_path":"CNN3d_ice_cube.ipynb","file_name":"CNN3d_ice_cube.ipynb","file_ext":"py","file_size_in_byte":3832,"program_lang":"python","lang":"en","doc_type":"code","stars":0,"dataset":"github-jupyter-script","pt":"40"}
+{"seq_id":"15299045054","text":"# +\nfrom strands import Schrodinger2D, Rectangle\n\ns = Schrodinger2D(lambda x, y: 10*(x*x + y*y), Rectangle(-1,1,-1,1), maxBasisSize=32, gridSize=(48,48))\n# -\n\neigs = s.eigenfunctions(10)\neigs\n\n# +\nimport numpy as np\nimport matplotlib.pyplot as plt\n\nE4, f4 = eigs[4]\nE5, f5 = eigs[5]\nprint(E4, E5)\nxs = np.linspace(-1,1, 1000)\nX, Y = np.meshgrid(xs, xs)\nZ = f4(X, Y) - 0.4*f5(X, Y)\n\nplt.pcolormesh(X, Y, Z)\n\n# +\nimport imageio\n\nimageio.imwrite('eigenfunction.exr', Z.astype(\"float32\"))\n# -\n\n\n","repo_name":"toonijn/PhD-thesis","sub_path":"flipbook/eigenfunction.ipynb","file_name":"eigenfunction.ipynb","file_ext":"py","file_size_in_byte":491,"program_lang":"python","lang":"en","doc_type":"code","stars":0,"dataset":"github-jupyter-script","pt":"40"}
+{"seq_id":"18920909408","text":"# + [markdown] id=\"view-in-github\" colab_type=\"text\"\n# \n\n# + id=\"aMLCuJS-WT6l\"\nimport matplotlib.pyplot as plt\nimport numpy as np\nimport pandas as pd\nimport datetime\nimport pywt\nfrom scipy.optimize import curve_fit\n\nimport seaborn as sns\nfrom sklearn.linear_model import LinearRegression\nfrom sklearn.preprocessing import PolynomialFeatures\nfrom scipy import stats\nimport scipy\n\n\n# + id=\"bBsBNF7VWYel\"\ndef denoise(y2):\n fft_signal = y2\n y3 = y2\n # y3 = []\n # count = 0\n # while(count < len(y2)):\n # y3.append(y2[count])\n # count += 1000\n def filter_signal(th):\n f_s = fft_filter(th)\n return np.real(np.fft.ifft(f_s))\n def fft_filter(perc):\n fft_signal = np.fft.fft(y3)\n fft_abs = np.abs(fft_signal)\n th=perc*(2*fft_abs[0:int(len(y3))]/((len(y3)))).max()\n fft_tof=fft_signal.copy()\n fft_tof_abs=np.abs(fft_tof)\n fft_tof[fft_tof_abs<=th]=0\n return fft_tof\n def fft_filter_amp(th):\n fft = np.fft.fft(y3)\n fft_tof=fft.copy()\n fft_tof_abs=np.abs(fft_tof)\n fft_tof_abs=2*fft_tof_abs/(len(fft_tof_abs)/2.)\n fft_tof_abs[fft_tof_abs<=th]=0\n return fft_tof_abs[0:int(len(fft_tof_abs)/2.)]\n th_list = np.linspace(0,1,5)\n th_list = th_list[0:len(th_list)-1]\n th_list = np.array([0, 0.25, 0.5, 0.75])\n th_list = np.linspace(0,0.02,1000)\n th_list = th_list[0:len(th_list)]\n p_values = []\n corr_values = []\n for t in th_list:\n filt_signal = filter_signal(t)\n res = stats.spearmanr(y3,y3-filt_signal)\n p_values.append(res.pvalue)\n corr_values.append(res.correlation)\n th_opt = th_list[np.array(corr_values).argmin()]\n opt_signal = filter_signal(th_opt)\n return opt_signal\n","repo_name":"ethanbrown33/enge_anc","sub_path":"DenoisingFunction.ipynb","file_name":"DenoisingFunction.ipynb","file_ext":"py","file_size_in_byte":1889,"program_lang":"python","lang":"en","doc_type":"code","stars":0,"dataset":"github-jupyter-script","pt":"40"}
+{"seq_id":"70084266362","text":"# # Banco de dados\n# Aqui nós criamos o banco de dados hm1. Ele conterá os dados coletados da página do autor desejado, utilizando a função scrape\n\nfrom sqlite3 import dbapi2 as sq3\nimport os\nimport sqlite3\nimport pandas as pd\nimport altair as alt\nimport networkx as nx\nimport matplotlib.pyplot as plt\nfrom scraper_scholar import * \n\n\ndados = scrape('Jeffrey Heer') #Coletando informações pela função scrape\nprint(dados[0:5])\n\n# Construindo o banco de dados\n\nourschema = '''\n DROP TABLE IF EXISTS \"author\";\n DROP TABLE IF EXISTS \"paper\";\n DROP TABLE IF EXISTS \"author_paper\";\n CREATE TABLE \"author\" (\n \"id\" INTEGER PRIMARY KEY NOT NULL ,\n \"author\" VARCHAR NOT NULL UNIQUE\n );\n CREATE TABLE \"paper\" ( \n \"id\" INTEGER PRIMARY KEY NOT NULL ,\n \"paper\" VARCHAR NOT NULL UNIQUE\n );\n CREATE TABLE \"author_paper\" (\n 'author_id' VARCHAR,\n 'paper_id' VARCHAR,\n FOREIGN KEY(author_id) REFERENCES author(id)\n FOREIGN KEY(paper_id) REFERENCES paper(id)\n );\n '''\n\nPATHSTART=\".\"\ndef get_db(dbfile):\n sqlite_db = sq3.connect(os.path.join(PATHSTART, dbfile))\n return sqlite_db\ndef init_db(dbfile, schema):\n \"\"\"Creates the database tables.\"\"\"\n db = get_db(dbfile)\n db.cursor()\n db.executescript(schema)\n db.commit()\n return db\n\n\ndb=init_db(\"hw1.db\", ourschema)\n\n# +\n#criando data papers, authors e authors_paper\nx = []\nfor i in dados:\n for j in i['authors']:\n x.append([i['title'],j.strip()])\nx = pd.DataFrame(x,columns=['paper','author'])\n\nx1 = list(sorted(set(x['paper'])))\nx2 = list(sorted(set(x['author'])))\npaper=pd.DataFrame(x1,columns=['paper'])\n\nauthor=pd.DataFrame(x2,columns=['author']) \n# -\n\npaper.head(5)\n\n# Inserindo dados\n\nauthor.to_sql(\"author\", db, if_exists=\"append\", index=False)\npaper.to_sql(\"paper\", db, if_exists=\"append\", index=False)\n\n\ndef make_query(tabela):\n c=db.cursor().execute(\"SELECT * FROM {};\".format(tabela))\n return c.fetchall()\n\n\n# +\na = make_query('author')\nt = make_query('paper')\na = pd.DataFrame(a,columns=['author_id','author'])\nt = pd.DataFrame(t,columns=['paper_id','paper'])\nauthor_paper = pd.merge(x,a,on='author') \nauthor_paper = pd.merge(author_paper,t,on='paper') \n\n\n# -\n\nauthor_paper.head(5)\n\n#author_paper = pd.concat([author_paper['author_id'],author_paper['paper_id']],axis=1)\nauthor_paper.to_sql(\"author_paper\", db, if_exists=\"append\", index=False,)\n\n\n","repo_name":"joaovictorpasseri/FDS-Homeworks","sub_path":"hm1/Banco de Dados e Site.ipynb","file_name":"Banco de Dados e Site.ipynb","file_ext":"py","file_size_in_byte":2425,"program_lang":"python","lang":"en","doc_type":"code","stars":0,"dataset":"github-jupyter-script","pt":"40"}
+{"seq_id":"41292385930","text":"# Our data management libraries\nimport pandas as pd\n\ndf = pd.read_csv('ecommerce-behavior-data-from-multi-category-store/2019-Nov.csv',chunksize = 10000)\n\nfor chunk in df:\n display(chunk.head())\n break\n\n\n","repo_name":"Hmmsien/ECommerce_behavior","sub_path":".ipynb_checkpoints/Ecommerce-Behavior-Analytics-checkpoint.ipynb","file_name":"Ecommerce-Behavior-Analytics-checkpoint.ipynb","file_ext":"py","file_size_in_byte":210,"program_lang":"python","lang":"en","doc_type":"code","stars":2,"dataset":"github-jupyter-script","pt":"40"}
+{"seq_id":"27490563458","text":"# + id=\"kPG2iGOUuinR\"\ndef scientificNotation(number, exponent):\n\n answer = 10**exponent\n answer = number * answer\n \n return answer\n\n\n\n# + [markdown] id=\"iqwkA_qSwaMa\"\n# 6\n\n# + colab={\"base_uri\": \"https://localhost:8080/\"} id=\"QEhpjiQRwKVW\" outputId=\"37a147fa-b3b4-4f36-8d4f-edb792d0bf64\"\nscientificNotation(4.9,-4) * scientificNotation(4.9,-4)\n\n# + colab={\"base_uri\": \"https://localhost:8080/\"} id=\"iXFw1ax2vHeo\" outputId=\"efca09ae-6281-45b1-b2d6-34b36086331b\"\nscientificNotation(-2.2, -4) / scientificNotation(2.4, -7)\n\n# + [markdown] id=\"XbsrrZBGwcok\"\n# 7\n\n# + colab={\"base_uri\": \"https://localhost:8080/\"} id=\"oQwOLr-NwdaX\" outputId=\"7620a0aa-2e54-4e90-a7b2-ddd0071bda32\"\nscientificNotation(5.3,-4) * scientificNotation(9.2,-4)\n\n# + colab={\"base_uri\": \"https://localhost:8080/\"} id=\"UhSLAxj3u-rL\" outputId=\"3b6f2786-3f94-4cb7-e4d1-8daebacb866a\"\nscientificNotation(1.8, -4) / scientificNotation(4.9, -7)\n\n# + [markdown] id=\"oE7bsn6lwqfF\"\n# 8\n\n# + colab={\"base_uri\": \"https://localhost:8080/\"} id=\"BkZ1f5VbwrVC\" outputId=\"3b278db8-983f-4443-8481-0573134ffe62\"\nscientificNotation(0.7,-4) * scientificNotation(-5.0,-4)\n\n# + colab={\"base_uri\": \"https://localhost:8080/\"} id=\"VuqKIaFlvNR-\" outputId=\"8286d162-553f-4d91-f07e-e0281a3a9502\"\nscientificNotation(6.4, -4) / scientificNotation(-3.5, -8)\n\n# + [markdown] id=\"gGjx8pQ3w23B\"\n# 9\n\n# + colab={\"base_uri\": \"https://localhost:8080/\"} id=\"MAagm4adw3wb\" outputId=\"ea9afad1-1040-413c-92b4-c1263080ca53\"\nscientificNotation(-1.0,-5) * scientificNotation(-5.3,-4)\n\n# + colab={\"base_uri\": \"https://localhost:8080/\"} id=\"C5_AjNtLvNbj\" outputId=\"4c9bfe71-9bfb-4e10-b44b-4d05952871b3\"\nscientificNotation(7.2, -4) / scientificNotation(-5, -9)\n\n# + [markdown] id=\"C01BiKzxxHLi\"\n# 10\n\n# + colab={\"base_uri\": \"https://localhost:8080/\"} id=\"5m3STG7_xKI8\" outputId=\"c1a2ceda-c9a4-4964-b4b5-10a5273a1bdf\"\nscientificNotation(3.9,-4) * scientificNotation(-1.1,-4)\n\n# + colab={\"base_uri\": \"https://localhost:8080/\"} id=\"5OFvg9NZvNjv\" outputId=\"1d6a974b-eeaa-4b14-e9d3-270da4de4779\"\nscientificNotation(-3.2, -4) / scientificNotation(-4.29, -8)\n","repo_name":"morganbenavidez/Chemistry","sub_path":"ScientificNotation.ipynb","file_name":"ScientificNotation.ipynb","file_ext":"py","file_size_in_byte":2076,"program_lang":"python","lang":"en","doc_type":"code","stars":0,"dataset":"github-jupyter-script","pt":"40"}
+{"seq_id":"36918986619","text":"# # Reinforcement learning\n# ### Snake - Monte Carlo\n# Play games and iteratively improve the state values and corresponding policy\n\nimport numpy as np\nimport random\nimport matplotlib.pyplot as plt\nfrom matplotlib.animation import FuncAnimation\nfrom matplotlib import animation\nfrom IPython.display import HTML, Image\n\n# ### Setup\n\n# +\n'''Modified with improved 'graphics' '''\n\nimport numpy as np\n\nim = 5\n\nclass Game():\n \n def __init__(self,grid,snake_size, snake_pad=2):\n self.board = np.ones((grid,grid,3))\n self.image = np.ones((grid*im,grid*im,3))\n self.snake = Snake(self.board,snake_size)\n self.pad = snake_pad\n self.apple = None\n self.score = 0\n self.total = 0\n self.deaths = 0\n\n def reset(self):\n self.snake.reset(self.pad)\n self.spawn_apple()\n\n def draw(self):\n # white background\n self.board[:,:,:] = 1\n self.image[:,:,:] = 1\n # black snake\n \n # head\n x,y = self.snake.pos\n self.board[x,y,:] = 0\n self.image[x*im:(x+1)*im,y*im:(y+1)*im,:] = 0\n d = self.snake.d\n xx,yy = [((x+1)*im-1,(y+1)*im-1),((x)*im,(y+1)*im-1),(x*im,y*im),(x*im,y*im)][d]\n self.image[xx,yy,:] = 1\n xx,yy = [((x+1)*im-1,y*im),((x+1)*im-1,(y+1)*im-1),(x*im,(y+1)*im-1),((x+1)*im-1,y*im)][d]\n self.image[xx,yy,:] = 1\n\n for s in self.snake.segments[1:]:\n x,y = s\n self.board[x,y,:] = 0\n self.image[x*im:(x+1)*im,y*im:(y+1)*im,:] = 0\n # tail\n if (len(self.snake.segments)>0):\n x,y = self.snake.segments[0]\n self.board[x,y,:] = 0 \n #self.image[x*im:(x+1)*im,y*im:(y+1)*im,1] = 0\n if (len(self.snake.segments)>1):\n n,m = self.snake.segments[1]\n else:\n n,m = self.snake.pos\n if (n==x and m>y):\n self.image[x*im:(x+1)*im,(y+1)*im-1,:] = 0\n self.image[x*im+1:(x+1)*im-1,(y+1)*im-2,:] = 0\n self.image[x*im+1:(x+1)*im-1,(y+1)*im-3,:] = 0\n self.image[x*im+2:(x+1)*im-2,(y+1)*im-4,:] = 0\n self.image[x*im+2:(x+1)*im-2,(y+1)*im-5,:] = 0\n elif (n==x and m\n\n# + id=\"uCaR8SOjNmfS\"\nimport numpy as np\n\n# + id=\"moeblV0gOS22\"\nfirst_numpy_array = np.array([1,2,3,4])\n\n# + colab={\"base_uri\": \"https://localhost:8080/\"} id=\"1lhudMRVOZxY\" outputId=\"d37f7840-e19e-423a-b3ed-f355600c09c0\"\nprint(first_numpy_array)\n\n# + colab={\"base_uri\": \"https://localhost:8080/\"} id=\"VT5LC5CBOcSX\" outputId=\"3150a2ee-6103-40f8-a330-450b08feaa86\"\narray_zero = np.zeros((3,3))\narray_zero\n\n# + colab={\"base_uri\": \"https://localhost:8080/\"} id=\"4HTctU22Ojqw\" outputId=\"5f8f1ea5-11b4-4f3d-a3e6-0002657f2032\"\narray_ones= np.ones((3,3))\narray_ones\n\n# + colab={\"base_uri\": \"https://localhost:8080/\"} id=\"-OgVVCAFOoUl\" outputId=\"6a1e4475-0d09-4678-c737-c4e33724e0ea\"\narray_empty = np.empty((2,3))\narray_empty\n\n# + colab={\"base_uri\": \"https://localhost:8080/\"} id=\"qnVflZsdOx_l\" outputId=\"9a5179a3-ae57-4290-98bc-dc176ee013b5\"\nnp_arange = np.arange(12)\nprint(np_arange)\n\n# + colab={\"base_uri\": \"https://localhost:8080/\"} id=\"ZjcWyC-TO3f-\" outputId=\"67956883-2fda-422e-f39c-62359218562e\"\nnp_arange.reshape(3,4)\n\n# + id=\"d2uFNunmPBKE\"\nnp_linspace = np.linspace(1,6,5)\n\n# + colab={\"base_uri\": \"https://localhost:8080/\"} id=\"AV7JGPrRPKCv\" outputId=\"172cc527-7abb-4ca5-d849-cf10c96046eb\"\nprint(np_linspace)\n\n# + colab={\"base_uri\": \"https://localhost:8080/\"} id=\"BdZ7HiXDPNOR\" outputId=\"2e230975-bb78-49ba-9984-ccd825885573\"\noneD_array = np.arange(15)\nprint(oneD_array)\n\n# + colab={\"base_uri\": \"https://localhost:8080/\"} id=\"Y_8pA6eUPcqd\" outputId=\"8d8a425b-0756-4226-867f-35b7621f057f\"\ntwoD_arr = oneD_array.reshape(3,5)\nprint(twoD_arr)\n\n# + colab={\"base_uri\": \"https://localhost:8080/\"} id=\"AtjNkVeqPi5U\" outputId=\"b176c0eb-13ec-4a33-93e4-5df1ed22d82d\"\nthreeD_array = np.arange(27).reshape(3,3,3)\nprint(threeD_array)\n\n# + id=\"YoVqzuyhPudV\"\n\n\n# + [markdown] id=\"onPa_VolfWRa\"\n# SciPy - Eigenvalues and Eigenvector\n\n# + id=\"bGoEPnb_fg7s\"\nimport numpy as np\nfrom scipy import linalg\n\n# + id=\"ByuV-m2Efx9w\"\ntest_rating_data = np.array([[5,8],[7,9]])\neigenValues, eigenVector = linalg.eig(test_rating_data)\nfirst_eigen, second_eigen = eigenValues\n\n# + colab={\"base_uri\": \"https://localhost:8080/\"} id=\"-RRZQxKAgR4F\" outputId=\"bb6efadb-8a3c-4fde-da74-b961ac86d7d7\"\nprint(first_eigen, second_eigen)\n\n# + colab={\"base_uri\": \"https://localhost:8080/\"} id=\"HTnIiGvFgYcl\" outputId=\"e8712207-dcba-495b-9f1b-6857f51900de\"\nprint(eigenVector[:,0])\n\n# + id=\"pFzzZzZfgfhr\"\n\n\n# + [markdown] id=\"Ppr0o7hPgsTn\"\n# SciPy - CDF and PDF\n\n# + id=\"0ZRMt8celdCS\"\nfrom scipy.stats import norm\n\n# + colab={\"base_uri\": \"https://localhost:8080/\"} id=\"pFFLcIpalgjb\" outputId=\"3b17d864-dc3c-4cfa-e408-2510522b0b3e\"\nnorm.rvs(loc=0,scale=1,size=20)\n\n# + colab={\"base_uri\": \"https://localhost:8080/\"} id=\"CC6tvKuslpMp\" outputId=\"9143d0d3-2ae1-4409-f089-712c1b951654\"\n#Cumulative Distribution Function\nnorm.cdf(10, loc=1,scale=3)\n\n# + colab={\"base_uri\": \"https://localhost:8080/\"} id=\"RzAyuQUYmc3V\" outputId=\"0b7182d8-ecfb-4fe9-8dae-48d76a7c0d8e\"\n#Probability Density Function\nnorm.pdf(14,loc=1,scale=1)\n\n# + id=\"IY63kXJkmorI\"\n\n\n# + id=\"nHP-InYRtrsf\"\n\n\n# + [markdown] id=\"xKzV-YTBtmUa\"\n# Web Scraping - Navigating the Tree\n\n# + id=\"bkEn62wntsbv\"\nfrom bs4 import BeautifulSoup\n\n# + id=\"Mt6Aak0FtzNb\"\nbook_html_doc = \"\"\n\n# + id=\"gsyspviTuGrk\"\n\n","repo_name":"khushboo-gehi/Data-Science-Projects-in-Python","sub_path":"_DSPy.ipynb","file_name":"_DSPy.ipynb","file_ext":"py","file_size_in_byte":3471,"program_lang":"python","lang":"en","doc_type":"code","stars":0,"dataset":"github-jupyter-script","pt":"40"}
+{"seq_id":"43015750568","text":"# +\n# Imports here\n# %matplotlib inline\n# %config InlineBackend.figure_format = 'retina'\n\nimport matplotlib.pyplot as plt\n\n### Import packages\nimport torch\nimport numpy as np\nfrom torch import nn, optim\nimport torch.nn.functional as F\nimport matplotlib.pyplot as plt\nfrom torchvision import datasets, transforms, models\nfrom PIL import Image\nimport os\n\n\n# -\n\ndef predict(image_path, model,device='cpu'):\n ''' Predict the class (or classes) of an image using a trained deep learning model.\n '''\n model.eval()\n eval_img = process_image(image_path)\n eval_img = eval_img.unsqueeze(0).to(device)\n logits = model.forward(eval_img)\n ps = F.softmax(logits, dim=1)\n return ps\n\n\n# +\nfrom collections import OrderedDict\ndef load_base_model(device, classes):\n model = models.resnet152(pretrained=True)\n\n ## Freeze parameters so we don't backprop through them\n for param in model.parameters():\n param.requires_grad = False\n \n for param in model.layer3.parameters():\n param.requires_grad = False\n \n for param in model.layer4.parameters():\n param.requires_grad = False\n\n \n fc = nn.Sequential(OrderedDict([\n ('fc1', nn.Linear(2048, 1024)),\n ('relu', nn.ReLU()),\n ('fc2', nn.Linear(1024, len(classes))),\n ('output', nn.LogSoftmax(dim=1))\n ]))\n\n model.fc = fc\n model.to(device)\n return model\n\ndef load_checkpoint(device, classes):\n model = load_base_model(device, classes)\n temp_model = torch.load('model_resnet_maxaccu_152_p99.pt', map_location=lambda storage, loc: storage)\n fc = nn.Sequential(OrderedDict([\n ('fc1', nn.Linear(2048, 1024)),\n ('relu', nn.ReLU()),\n ('fc2', nn.Linear(1024, len(temp_model['fc.fc2.weight']))),\n ('output', nn.LogSoftmax(dim=1))\n ]))\n model.fc = fc\n model.load_state_dict(torch.load('model_resnet152_p99.pt', map_location=lambda storage, loc: storage))\n return model\n\n\n# +\ndef imshow(image, ax=None, title=None, normalize=True):\n \"\"\"Imshow for Tensor.\"\"\"\n if ax is None:\n fig, ax = plt.subplots()\n image = image.numpy().transpose((1, 2, 0))\n\n if normalize:\n mean = np.array([0.485, 0.456, 0.406])\n std = np.array([0.229, 0.224, 0.225])\n image = std * image + mean\n image = np.clip(image, 0, 1)\n\n ax.imshow(image)\n ax.spines['top'].set_visible(False)\n ax.spines['right'].set_visible(False)\n ax.spines['left'].set_visible(False)\n ax.spines['bottom'].set_visible(False)\n ax.tick_params(axis='both', length=0)\n ax.set_xticklabels('')\n ax.set_yticklabels('')\n\n return ax\n\ndef top_classes(img_path, model, device, classes):\n processed_image = process_image(img_path)\n img_result = predict(img_path, model, device)\n probs, idx = torch.topk(img_result, 10)\n probs,idx = probs.squeeze(), idx.squeeze()\n \n title = img_path\n img = Image.open(img_path)\n \n labels = []\n probabilities = []\n for i,prob in enumerate(probs):\n labels.append(classes[idx[i].item()])\n probabilities.append(prob.item())\n print(img_result.size())\n \n fig, (ax1, ax2) = plt.subplots(figsize=(6,10), ncols=1, nrows=2)\n ax1.set_title(title)\n ax1.imshow(img)\n ax1.axis('off')\n y_pos = np.arange(len(probabilities))\n ax2.barh(y_pos, probabilities)\n ax2.set_yticks(y_pos)\n ax2.set_yticklabels([x for x in labels])\n ax2.invert_yaxis()\n \n \ndef process_image(image):\n ''' Scales, crops, and normalizes a PIL image for a PyTorch model,\n returns an Numpy array\n '''\n \n # TODO: Process a PIL image for use in a PyTorch model\n img = Image.open( image )\n img.load()\n normalize = transforms.Normalize(\n mean=[0.485, 0.456, 0.406],\n std=[0.229, 0.224, 0.225]\n )\n preprocess = transforms.Compose([\n transforms.Resize(256),\n transforms.CenterCrop(224),\n transforms.ToTensor(),\n normalize\n ])\n img_tensor = preprocess(img)\n return img_tensor\n\ndef getClassesFromFiles():\n my_classes = []\n for filename in os.listdir(\"./url_files/\"):\n if filename.find(\"_valid.txt\") == -1 and filename.find(\".txt\") > 0:\n my_classes.append(filename.replace('.txt',''))\n my_classes = sorted(my_classes)\n return my_classes\n\n\n# -\n\ndevice_to = torch.device(\"cuda\" if torch.cuda.is_available() else \"cpu\")\nclass_names = getClassesFromFiles()\ntop_classes('data/valid/autumn/00000006.jpg', load_checkpoint(device_to, class_names), device_to, class_names)\n\n# !wget https://images.pexels.com/photos/1633522/pexels-photo-1633522.jpeg\nprint(device_to)\n\n\n# !mv pexels-photo-1633522.jpeg running-dogs.jpeg\n\ntop_classes('running-dogs.jpeg', load_checkpoint(device_to, class_names), device_to, class_names)\n","repo_name":"ismaproco/image2hashtag","sub_path":"predict_classes.ipynb","file_name":"predict_classes.ipynb","file_ext":"py","file_size_in_byte":4970,"program_lang":"python","lang":"en","doc_type":"code","stars":1,"dataset":"github-jupyter-script","pt":"40"}
+{"seq_id":"11603229766","text":"# # Exercici 1\n# #### Descarrega el dataset adjunt preu lloguer per trimestre i barri.csv extret de la web OpenDataBCN i resumeix-lo estadísticament i gràficament.\n#\n# #### Crea almenys una visualització per:\n\n# +\nimport numpy as np\nimport pandas as pd\n\nimport matplotlib as mpl\nimport matplotlib.pyplot as plt\nimport seaborn as sns\n# -\n\ndf = pd.read_csv('C:/RAUL/NISSAN/GENERAL/USB/FORMACION/BootCamp - Data Scientist_Analist/IT Academy Cibernarium - Data Science/Data Science Itinerario/4.- Visualitzación gráfica de datos/Entrega/preu lloguer per trimestre i barri.csv')\n\n\n# * Resum estadístic\n\ndf.isnull().sum()\n\ndf.head()\n\ndf['Preu'] = df['Preu'].astype(float) # Canviem el tipus de la columna preu per poder fer càlculs\n\n#Eliminem els registres que no ténen dades de Preu ja que a priori no són interessants per fer l'anàlisis\ndf = df[df['Preu'] != '--']\n\ndf.shape\n\ndf.info()\n\ndf_Preu_mensual = df[df['Lloguer_mitja'] == 'Lloguer mitjà mensual (Euros/mes)']\ndf_Preu_mensual.describe().round(4)\n\ndf_Preu_m2 = df[df['Lloguer_mitja'] == 'Lloguer mitjà per superfície (Euros/m2 mes)']\ndf_Preu_m2.describe().round(4)\n\ndf_pvt=df[[\"Nom_Districte\", \"Lloguer_mitja\"]]\npivotTable = df_pvt.pivot_table(index= \"Nom_Districte\", columns= \"Lloguer_mitja\", values=\"Lloguer_mitja\", aggfunc= 'size', fill_value=0)\npivotTable\n\n# * Una variable categòrica (Districte o Barri)\n\ndf_Barris=df['Nom_Districte']\ndf_Barris = df_Barris.value_counts().rename_axis('Barris').reset_index(name='N_registres')\ndf_Barris.set_index('Barris', inplace=True) #Quan es dibuixa directament amb plotly l'índex és el que es fa servir per\n# les categories de les coordenades X, si es fa amb el \n\n# +\np1=df_Barris.plot(kind='bar', figsize=(8, 5))\nplt.title('Registres per districtes')\nplt.ylabel('N_registres')\nplt.xlabel('Districtes')\nplt.figure(figsize=(10, 2.7))\n\nplt.show()\n# -\n\n# * Una variable numèrica (Preu)\n\ndf_Preu_mensual_1=df_Preu_mensual.sort_values(by='Preu', ascending=False)\ndf_Preu_mensual_1.reset_index(drop=True, inplace=True)\ndf_Preu_mensual_1=df_Preu_mensual_1['Preu']\n\ndf_Preu_m2_1=df_Preu_m2.sort_values(by='Preu', ascending=False)\ndf_Preu_m2_1.reset_index(drop=True, inplace=True)\ndf_Preu_m2_1=df_Preu_m2_1['Preu']\n\n# +\nplt.subplot(1, 2, 1)\n\nplt.subplot(1, 2, 1)\np2=df_Preu_mensual_1.plot(kind='hist', x='df_Preu_mensual', figsize=(10, 6))\nplt.title('Histograma Preus mensual')\nplt.xlabel('€ / mes')\nplt.ylabel('Registres')\n\nplt.subplot(1, 2, 2)\np3=df_Preu_m2_1.plot(kind='hist', x='df_Preu_m2', figsize=(8, 6))\nplt.title('Histograma Preus per m2')\nplt.xlabel('€ / m2')\nplt.ylabel('Registres')\n# -\n\n# * Una variable numèrica i una categòrica (Districte i Preu)\n\ndf_Preu_mensual_2=df_Preu_mensual.sort_values(by='Preu', ascending=False)\ndf_Preu_mensual_2.reset_index(drop=True, inplace=True)\ndf_Preu_mensual_Districte=df_Preu_mensual_2[['Nom_Districte','Preu']]\n\ndf_Preu_m2_2=df_Preu_m2.sort_values(by='Preu', ascending=False)\ndf_Preu_m2_2.reset_index(drop=True, inplace=True)\ndf_Preu_m2_Districte=df_Preu_m2_2[['Nom_Districte','Preu']]\n\n# +\nplt.subplot(1, 2, 1)\n\n# Adjust the subplot layout parameters\n\nplt.subplot(1, 2, 1)\np4=plt.scatter(df_Preu_mensual[\"Nom_Districte\"].astype(str), df_Preu_mensual[\"Preu\"])\nplt.margins(x=0.1)\nplt.xticks(rotation=90)\n\nplt.title('Preus mensuals per distrticte', y=1.05)\nplt.xlabel('Districtes')\nplt.ylabel('€ / mes')\n\n\n\nplt.subplot(1, 2, 2)\np5=plt.scatter(df_Preu_m2[\"Nom_Districte\"].astype(str), df_Preu_m2[\"Preu\"])\nplt.margins(x=0.1)\nplt.xticks(rotation=90)\n\nplt.title('Preus per m2 i distrticte', y=1.05)\nplt.xlabel('Districtes')\nplt.ylabel('€ / m2')\n\nplt.subplots_adjust(right=1.50)\n\nplt.show()\n\n# +\nf, axes = plt.subplots(1, 2, figsize=(12, 5))\n\nax0=sns.stripplot(data=df_Preu_mensual, x=\"Nom_Districte\", y=\"Preu\", ax=axes[0])\nax0.set_xticklabels(ax0.get_xticklabels(),rotation=90)\nax0.set(xlabel='Districtes', ylabel='€ / mes')\nax0.set_title(\"Preus mensuals per distrticte\")\n\nax1=sns.stripplot(data=df_Preu_m2, x=\"Nom_Districte\", y=\"Preu\", ax=axes[1])\nax1.set_xticklabels(ax1.get_xticklabels(),rotation=90)\nax1.set(xlabel='Districtes', ylabel='€ / m2')\nax1.set_title(\"Preus m2 per distrticte\") \n\n\n# +\nf, axes = plt.subplots(1, 2, figsize=(12, 5))\n\nax2=sns.boxplot(data=df_Preu_mensual, x=\"Nom_Districte\", y=\"Preu\", ax=axes[0])\nax2.set_xticklabels(ax0.get_xticklabels(),rotation=90)\nax2.set(xlabel='Districtes', ylabel='€ / mes')\nax2.set_title(\"Preus mensuals per distrticte\")\n\nax3=sns.boxplot(data=df_Preu_m2, x=\"Nom_Districte\", y=\"Preu\", ax=axes[1])\nax3.set_xticklabels(ax1.get_xticklabels(),rotation=90)\nax3.set(xlabel='Districtes', ylabel='€ / m2')\nax3.set_title(\"Preus m2 per distrticte\") \n\n# -\n\n# * Dues variables numèriques (Any o Trimestre i Preu)\n\n# +\nf, axes = plt.subplots(1, 2, figsize=(10, 4))\n\nax4=sns.violinplot(data=df_Preu_mensual, x=\"Trimestre\", y=\"Preu\", ax=axes[0], inner=None, color=\".8\")\nax4=sns.stripplot(data=df_Preu_mensual, x=\"Trimestre\", y=\"Preu\", ax=axes[0])\nax4.set(xlabel='Trimestre', ylabel='€ / mes')\nax4.set_title(\"Preus mensuals per trimestre\")\n\nax5=sns.violinplot(data=df_Preu_m2, x=\"Trimestre\", y=\"Preu\", ax=axes[1], inner=None, color=\".8\")\nax5=sns.stripplot(data=df_Preu_m2, x=\"Trimestre\", y=\"Preu\", ax=axes[1])\nax5.set(xlabel='Trimestre', ylabel='€ / m2')\nax5.set_title(\"Preus m2 per trimestre\") \n# -\n\n# * Tres variables (Barri o Districte, Trimestre i Preu)\n\n# +\nf, axes = plt.subplots(1, 2, figsize=(20, 8))\n\nax6=sns.barplot(data=df_Preu_mensual, x=\"Nom_Districte\", y=\"Preu\", ax=axes[0], hue='Trimestre')\nax6.set_xticklabels(ax0.get_xticklabels(),rotation=90)\nax6.set_ylim(500, 1900)\nax6.set(xlabel='Districtes', ylabel='€ / mes')\nax6.set_title(\"Preus mensuals per distrticte i trimestre\")\n\nax7=sns.barplot(data=df_Preu_m2, x=\"Nom_Districte\", y=\"Preu\", ax=axes[1], hue='Trimestre')\nax7.set_xticklabels(ax1.get_xticklabels(),rotation=90)\nax7.set_ylim(8, 18)\nax7.set(xlabel='Districtes', ylabel='€ / m2')\nax7.set_title(\"Preus m2 per distrticte i trimestre\") \n\n\n# -\n\n# # Exercici 2\n# #### Exporta els gràfics com imatges o com HTML.\n\n# Funció per exportar gràfics a partir del nom que li volem donar i el que té asignat.\ndef exporta_a_jpg(nombre_fichero, grafico):\n fig = grafico.get_figure()\n fig.savefig(nombre_fichero+\".jpeg\")\n\n\nexporta_a_jpg('Registres per districtes', p1)\n\nexporta_a_jpg('Histograma Preus mensual + Preus m2', p2)\n\nexporta_a_jpg('Preus mensuals per distrticte i preus m2 per districte', p4)\n\nexporta_a_jpg('Preus mensuals per distrticte i preus m2 per districte_2', ax0)\n\nexporta_a_jpg('Preus mensuals per distrticte i preus m2 per districte_2 (Boxplot)', ax2)\n\nexporta_a_jpg('Preus mensuals per trimestre i preus m2 per trimestre (Violinplot)', ax4)\n\nexporta_a_jpg('Preus mensuals i preus m2 per districte i trimestre (Barplot)', ax6)\n\n# # Exercici 3\n# #### Proposa alguna visualització que creguis que pot resultar interessant.\n\n# +\nfig, ax = plt.subplots(figsize=(10, 5))\n\ndf_Barris.plot.pie(y='N_registres', legend=False, autopct='%1.2f%%', ax=ax)\nplt.title('Registres per districtes')\nax.set_ylabel(None)\nplt.show()\n\n# -\n\ndf_Preu_mensual_2=df_Preu_mensual[[\"Nom_Districte\", \"Preu\"]]\n\nfig, ax = plt.subplots(figsize=(10,6))\nsns_plot = sns.histplot(data=df_Preu_mensual_2, x=\"Preu\", hue=\"Nom_Districte\", multiple=\"stack\")\nax.set_title(\"Preus mensuals per distrticte\")\n\ndf_Preu_mensual_2=df_Preu_mensual[[\"Nom_Districte\", \"Preu\"]]\n\n# +\ndf_Preu_m2_2=df_Preu_m2[[\"Nom_Districte\", \"Preu\"]]\n\nfig, ax = plt.subplots(figsize=(10,6))\nsns.histplot(data=df_Preu_m2_2, x=\"Preu\", hue=\"Nom_Districte\", multiple=\"stack\")\nax.set_title(\"Preus m2 per distrticte\")\n","repo_name":"raugaca/estructures_Dataframe_Sprint4","sub_path":"Tasca M4 T01.ipynb","file_name":"Tasca M4 T01.ipynb","file_ext":"py","file_size_in_byte":7603,"program_lang":"python","lang":"en","doc_type":"code","stars":0,"dataset":"github-jupyter-script","pt":"4"}
+{"seq_id":"17356367801","text":"# # Amortization (workspace notebook)\n#\n# We can use this notebook to work on our `amortization` application. Since this is located on `src/classworks` we can import directly the `amortization` module.\n\n# We use the `settings` module to save the applications configuration. In this case, we can get the `AMORTIZATION_TABLE_DIRPATH` that represents the \"location\" of the `TABLE` directory in the filesystem.\n#\n# Let's create the folder if not exists. We can add this logic into the `setup` command:\n\nimport os\nfrom amortization.settings import AMORTIZATION_TABLE_DIRPATH\n\nAMORTIZATION_TABLE_DIRPATH\n\n# Create the folder if not exists\nif not os.path.exists(AMORTIZATION_TABLE_DIRPATH):\n os.makedirs(AMORTIZATION_TABLE_DIRPATH)\n\n# Import the `Amortization` class and create two instances.\n\n# +\nimport pandas as pd\nfrom fintools import Amortization\n\npd.options.display.float_format = '${:,.2f}'.format\n\n# Create a couple Amortization instances\namortization_a = Amortization(amount=50000, rate=0.08, n=8)\namortization_b = Amortization(amount=18000, rate=0.34/12, n=6)\n# -\n\n# We should be able to get a dataframe\ndf = amortization_b.get_table()\ndf\n\n# We should be able to get other implementations we already have on fintools:\n\n# +\nfrom fintools.rates import Cetes28\n\ncetes28 = Cetes28()\n# -\n\ncetes28.get_dataframe().iloc[-1].value\n\n# Let's create a quick visualization:\n\nimport matplotlib.pyplot as plt\n\n# +\na = Amortization(amount=18000, rate=0.10, n=10)\ndf = a.get_table()\n\nplot = df.plot.bar(x=\"t\", y=[\"principal\", \"interest\"], stacked=True)\nplt.title(\"Amortization Payments: ${:,.2f}\".format(a.annuity))\nplt.ylabel(\"$$$\")\nplt.grid()\nplt.show()\n# -\n\ndf.annuity[1:].unique().tolist().pop()\n\nplot.get_figure()\n\n\n","repo_name":"douglasparism/CREDIT-RISK-2021A","sub_path":"src/classwork/amortization.ipynb","file_name":"amortization.ipynb","file_ext":"py","file_size_in_byte":1711,"program_lang":"python","lang":"en","doc_type":"code","stars":0,"dataset":"github-jupyter-script","pt":"4"}
+{"seq_id":"699145294","text":"# \n\n# normalization\n# 适用于分布有明显便捷的情况; 受outlier影响较大\n# eg:学生分数, 像素点\n#\n# 将所有的数据映射到统一尺度\n# 最值归一化: 把所有数据映射到0-1之间\n# \n\n# ## 均值方差归一化 standardization\n# ### 把所有数据归一到均值为0方差为1的分布中, 适用于数据分布没有明显的边界, 有可能存在极端数据值\n\n# \n\n# # 数据归一化处理\n\nimport numpy as np\nimport matplotlib.pyplot as plt\n\n# ## 最值归一化 Normalization\n\nx = np.random.randint(0, 100, size=100)\n\nx\n\n(x - np.min(x)) / (np.max(x)- np.min(x))\n\nX = np.random.randint(0, 100, (50, 2))\n\nX[:10, :]\n\nX = np.array(X, dtype=float)\n\nX[:10, :]\n\nX[:, 0] = (X[:, 0] - np.min(X[:,0])) / (np.max(X[:,0]) - np.min(X[:,0]))\nX[:, 1] = (X[:, 1] - np.min(X[:,1])) / (np.max(X[:,1]) - np.min(X[:,1]))\n\nX[:10, :]\n\nplt.scatter(X[:, 0], X[:,1])\nplt.show()\n\nnp.mean(X[:,0])\n\nnp.mean(X[:,1])\n\nnp.std(X[:, 0])\n\n# ## 均值方差归一化 Standardization\n\nX2 = np.random.randint(0, 100, (50, 2))\n\nX2 = np.array(X2, dtype=float)\n\nX2[:10,:]\n\nX2[:,0] = (X2[:,0] - np.mean(X2[:,0])) / np.std(X2[:, 0])\n\nX2[:,1] = (X2[:,1] - np.mean(X2[:,1])) / np.std(X2[:, 1])\n\nX2[:10, :]\n\nplt.scatter(X2[:,0], X2[:,1])\nplt.show()\n\nnp.mean(X2[:,0])\n\nnp.std(X2[:,0])\n\n\n","repo_name":"DorianXXUUU/MachineLearningNotebook","sub_path":"KNN/06-Feature_Scaling.ipynb","file_name":"06-Feature_Scaling.ipynb","file_ext":"py","file_size_in_byte":1384,"program_lang":"python","lang":"en","doc_type":"code","stars":1,"dataset":"github-jupyter-script","pt":"4"}
+{"seq_id":"16992545236","text":"import pydotplus\nfrom sklearn.tree import DecisionTreeRegressor\nfrom sklearn.linear_model import LinearRegression\nfrom sklearn.metrics import mean_squared_error, accuracy_score\nfrom sklearn.base import BaseEstimator\nimport numpy as np\nimport math\nimport matplotlib.pyplot as plt\nfrom IPython.display import Image\nfrom sklearn import tree\nfrom scipy import optimize\nfrom sklearn import datasets, cross_validation, metrics, neighbors\nfrom matplotlib.colors import ListedColormap\nfrom pandas import DataFrame\n# %pylab inline\n\nwith open('german_credit.csv') as f:\n x = [i for i in f]\n\nfor i in range(len(x)):\n if i > 0:\n x[i] = (x[i].split(','))\ny = x[0]\nx = x[1:]\ntargets = []\nfor i in range(len(x)):\n for j in range(len(x[i]) - 1):\n x[i][j] = int(x[i][j])\n x[i][len(x[i]) - 1] = int(x[i][len(x[i]) - 1][0])\n targets.append(x[i][0])\n x[i] = x[i][1:]\n\nx = np.array(x)\n\ny = y.split(',')\n\ny[9] = y[9].replace('&', '')\n\ny\n\nmodel = tree.DecisionTreeClassifier(max_depth=3)\nmodel.fit(x, targets)\n\ndot_data = tree.export_graphviz(model, out_file=\"small_tree.out\", \n feature_names=y[1:],\n max_depth = 3,\n class_names=['0', '1'], \n filled=True, rounded=True, \n special_characters=True)\n\ngraph = pydotplus.graphviz.graph_from_dot_file(\"small_tree.out\") \nImage(graph.create_png())\n\n# #### Дерево наилучшим образом разбивает выборку по признакам. В первом узле разбиение выполнено по признаку, отвечающему за баланс пользователя. Причем видно, что в левой части появляются узлы, в которых алгоритм отвечает нулем, то есть кредит не выдается. Что неудивительно, так как баланс клиента низкий. То же самое можно сказать про длительность неоплаченного кредита и т.д.\n\ndepth_of_tree = np.linspace(1, 100, 100)\nac_cross = [0.0 for i in range(len(depth_of_tree))]\nac_train = [0.0 for i in range(len(depth_of_tree))]\nfor k in range(1, len(depth_of_tree)):\n model = tree.DecisionTreeClassifier(max_depth=k)\n model.fit(x, targets)\n ac_cross[k] = cross_validation.cross_val_score(model, x, targets).mean()\n ac_train[k] = accuracy_score(targets, model.predict(x))\n\npyplot.plot(depth_of_tree[1:], ac_cross[1:], color = 'red', label = 'cross_validation')\npyplot.plot(depth_of_tree[1:], ac_train[1:], color = 'blue', label = 'train_dataset')\npyplot.xlabel(\"k\")\npyplot.ylabel(\"accuracy\")\npyplot.title(\"accuracy\")\naxis = pyplot.gca()\naxis.legend(loc = 4)\npyplot.ylim(0, 1.1)\n\n# ### На обучающей выборке качество быстро выходит на 1. А как же иначе? Ведь она обучающая. Алгоритм подогнался, и при достаточно большой глубине дерева выдает точные ответы. На кроссвалидации точность примерно 0.7 при всех k.\n\n\n","repo_name":"iat7/ML2017","sub_path":"hw_2_2.ipynb","file_name":"hw_2_2.ipynb","file_ext":"py","file_size_in_byte":3228,"program_lang":"python","lang":"en","doc_type":"code","stars":0,"dataset":"github-jupyter-script","pt":"4"}
+{"seq_id":"24113957090","text":"# Modelo para anlisis de sentmientos para luego precedir si tiene sintomas de depresion.\n\nfrom string import punctuation\nimport pandas as pd\nimport nltk\nimport re\n\n# read the data from tweets_public.csv and create a dataframe\ndf = pd.read_csv('data/tweets_public.csv', encoding='utf-8')\ndf.head()\n\ndf.shape\n\n# # Preprocesamiento\n# Limpieza de datos\n\n# create a new dataframe with only the text and airline_sentiment columns and tweet id with the name df_sentiment\ndf_sentiment = df[['text', 'airline_sentiment', 'tweet_id']]\n\n# transform the text letters to lowercase\ndf_sentiment['text'] = df_sentiment['text'].str.lower()\ndf_sentiment.head()\n\n# Referencia: https://stackoverflow.com/questions/6718633/python-regular-expression-again-match-url\n# remove the urls from the text but keep all the text after the url\ndf_sentiment.loc[:, 'text'] = df_sentiment['text'].apply(lambda x: re.split('http[s]*\\S+', str(x))[0])\ndf_sentiment.head()\n\n# remove the punctuation from the text\ndf_sentiment.loc[:, 'text'] = df_sentiment['text'].apply(lambda x: ''.join(c for c in x if c not in punctuation))\ndf_sentiment.head()\n\n# change the \\n to a space\ndf_sentiment.loc[:, 'text'] = df_sentiment['text'].apply(lambda x: x.replace('\\n', ' '))\ndf_sentiment.head()\n\n# +\n# remove the stopwrods from the text\nnltk.download('stopwords')\nfrom nltk.corpus import stopwords\n\nstop_words = set(stopwords.words('spanish'))\ndf_sentiment.loc[:, 'text'] = df_sentiment['text'].apply(lambda x: ' '.join([word for word in x.split() if word.lower() not in stop_words]))\ndf_sentiment.head()\n\n# +\n# remove the emojis from the text\n\n# regular expression pattern to remove emojis from text\nemoji_pattern = re.compile(\"[\"\n u\"\\U0001F600-\\U0001F64F\" # faces\n u\"\\U0001F300-\\U0001F5FF\" # simbols & pictographs\n u\"\\U0001F680-\\U0001F6FF\" # transport & map symbols\n u\"\\U0001F1E0-\\U0001F1FF\" # flags\n \"]+\", flags=re.UNICODE)\n\n# use lambda function to remove the emojis from the text\ndf_sentiment.loc[:, 'text'] = df_sentiment['text'].apply(lambda x: emoji_pattern.sub(r'', x))\ndf_sentiment.head()\n# -\n\n# remove the numbers from the text\ndf_sentiment.loc[:, 'text'] = df_sentiment['text'].apply(lambda x: re.sub(r'\\d+', '', x))\ndf_sentiment.head()\n\ndf_sentiment = df[['text', 'airline_sentiment']]\n# df_sentiment['label'] = df_sentiment['airline_sentiment'].apply(lambda x: '1 start' if x == 'negative' else '5 stars')\ndf_sentiment['label'] = df_sentiment['airline_sentiment'].apply(lambda x: 0 if x == 'negative' else 1)\ndf_sentiment = df_sentiment[['text', 'label']]\ndf_sentiment\n\n# save the dataframe to a csv file\ndf_sentiment.to_csv('data/clean_tweets_sentiment.csv', index=False)\n\n# # Hugging Face\n\nmodel_preload1 = \"nlptown/bert-base-multilingual-uncased-sentiment\"\n# model_preload = \"distilbert-base-uncased\"\n# model_preload = \"xlm-roberta-base\"\n# model_preload = \"distilbert-base-multilingual-cased\"\n# model_preload = \"bert-base-multilingual-cased\"\nmodel_preload = 'dccuchile/bert-base-spanish-wwm-cased'\n# model_preload = \"RoBERTa-base\"\n\n# +\n\ndata = [\"Amo el mundo en el que vivo\", \"Odio a todas las personas que conozco\"]\n# -\n\nfrom transformers import pipeline\nsentiment_pipeline = pipeline(\"sentiment-analysis\")\nsentiment_pipeline(data)\n\nsentiment_pipeline('no estoy normal')\n\nspecific_model = pipeline(\"sentiment-analysis\", model=model_preload1)\nspecific_model(data)\n\nspecific_model('estoy normal')\n\nspecific_model('estoy triste')\n\n# ## Fine tuning\n\n# +\n# import torch\n# torch.cuda.is_available()\n\n# +\n# torch.cuda.empty_cache()\n# -\n\nfrom datasets import load_dataset\nlocal_train = load_dataset(\"csv\", data_files=\"data/clean_tweets_sentiment.csv\", split=\"train[:70%]\")\nlocal_test = load_dataset(\"csv\", data_files=\"data/clean_tweets_sentiment.csv\", split=\"train[70%:]\")\n\nprint(local_train.shape)\nprint(local_test.shape)\nprint(local_train[0])\nprint(local_test[0])\n\n# +\nsmall_train_dataset = local_train.shuffle(seed=42).select(\n list(list(range(300)))\n)\nsmall_test_dataset = local_test.shuffle(seed=42).select(list(list(range(100))))\n# # split df_sentiment in 70% train and 30% test\n# from sklearn.model_selection import train_test_split\n\n# train, test = train_test_split(df_sentiment, test_size=0.3, random_state=42)\n# train.shape, test.shape\n# -\n\nfrom transformers import AutoTokenizer\ntokenizer = AutoTokenizer.from_pretrained(model_preload)\n\n\n# +\ndef preprocess_function(examples):\n return tokenizer(examples[\"text\"], truncation=True)\n\n# user the function preprocess_function to tokenize the train and test data\n# tokenized_train = train['text'].apply(lambda x: tokenizer(x, truncation=True))\n# tokenized_test = test['text'].apply(lambda x: tokenizer(x, truncation=True))\ntokenized_train = small_train_dataset.map(preprocess_function, batched=True)\ntokenized_test = small_test_dataset.map(preprocess_function, batched=True)\n# -\n\nfrom transformers import DataCollatorWithPadding\ndata_collator = DataCollatorWithPadding(tokenizer=tokenizer)\n\n# ## train the model\n\nfrom transformers import AutoModelForSequenceClassification\nmodel = AutoModelForSequenceClassification.from_pretrained(model_preload, num_labels=2, id2label={0: \"negative\", 1: \"positive\"})\n\n\n# +\nimport numpy as np\nfrom datasets import load_metric\n \ndef compute_metrics(eval_pred):\n load_accuracy = load_metric(\"accuracy\")\n load_f1 = load_metric(\"f1\")\n \n logits, labels = eval_pred\n predictions = np.argmax(logits, axis=-1)\n accuracy = load_accuracy.compute(predictions=predictions, references=labels)[\"accuracy\"]\n f1 = load_f1.compute(predictions=predictions, references=labels)[\"f1\"]\n return {\"accuracy\": accuracy, \"f1\": f1}\n\n\n# -\n\nfrom huggingface_hub import notebook_login\nnotebook_login()\n\n# +\nfrom transformers import TrainingArguments, Trainer\n \nrepo_name = \"raulangelj/huggingface_sentiment_analysis\"\n \ntraining_args = TrainingArguments(\n output_dir=repo_name,\n learning_rate=2e-5,\n per_device_train_batch_size=25,\n per_device_eval_batch_size=25,\n num_train_epochs=2,\n weight_decay=0.01,\n save_strategy=\"epoch\",\n push_to_hub=True,\n)\n \ntrainer = Trainer(\n model=model,\n args=training_args,\n train_dataset=tokenized_train,\n eval_dataset=tokenized_test,\n tokenizer=tokenizer,\n data_collator=data_collator,\n compute_metrics=compute_metrics,\n)\n# -\n\ntrainer.train()\n\ntrainer.evaluate()\n\ntrainer.push_to_hub()\n\ndata\n\n# +\nfrom transformers import pipeline\n \nsentiment_model = pipeline(model=repo_name)\nsentiment_model(data)\n# -\n\nsentiment_model('no me siento bien')\n\nsentiment_model('hoy es el mejor dia de mi vida')\n\nsentiment_model('ya no puedo con mi vida, no quiero que me hablen')\n\n# # Funcion a utilizar en proyecto\n\n# +\nimport unicodedata\nfrom transformers import pipeline\n\ndef has_first_person_pronouns(text):\n first_person_pronouns = [\"yo\", \"me\", \"mi\", \"conmigo\"]\n \n words = text.split()\n for word in words:\n normalize_word = remove_accent(word.lower())\n if normalize_word in first_person_pronouns:\n return True\n \n return False\n\ndef remove_accent(word):\n return ''.join((c for c in unicodedata.normalize('NFD', word) if unicodedata.category(c) != 'Mn'))\n\n# # ? should we use this?\ndef has_depression_words(text):\n depression_words = [\n \"depresion\", \"tristeza\", \"desesperanza\", \"desmotivacion\", \"soledad\", \"ansiedad\",\n \"desgano\", \"desanimo\", \"desaliento\", \"abatimiento\", \"melancolia\", \"desolacion\",\n \"apatia\", \"angustia\", \"pesimismo\", \"desesperacion\", \"desamparo\", \"desconsuelo\",\n \"agotamiento\", \"cansancio\", \"desinteres\", \"insomnio\", \"culpa\", \"llanto\", \"suicidio\",\n \"autolesion\", \"desorden\", \"trastorno\", \"descontrol\", \"vacio\", \"apetito\", \"fatiga\",\n \"inutilidad\", \"retroceso\", \"preocupacion\", \"retraimiento\", \"negatividad\", \"desvalorizacion\",\n \"desesperado\", \"desesperanza\", \"perdida\", \"esperanza\", \"autoestima\", \"autoestima\",\n \"irritabilidad\", \"decision\", \"concentracion\", \"motivacion\", \"indefension\", \"vida\",\n \"conexion\", \"pena\", \"abandono\", \"inseguridad\", \"desapego\", \"agonia\", \"miedo\", \"rumiacion\",\n \"tratamiento\", \"terapia\", \"psicologo\", \"psiquiatra\", \"medicacion\", \"cura\", \"superacion\",\n \"autoayuda\", \"apoyo\", \"rehabilitacion\"\n ]\n \n words = text.split()\n for word in words:\n normalize_word = remove_accent(word.lower())\n if normalize_word in depression_words:\n return True\n \n return False\n\ndef has_possible_depression_sintom(text):\n # To know if has depression sintoms, we will check 3 things:\n # 1. If the text has first person pronouns\n # 2. If the text has negative sentiment\n # 3. If the text has words related to depression\n # If the text has the 3 things, we will say that the text has depression sintoms\n # If the text has 2 of the 3 things, we will say that the text has possible depression sintoms\n # If the text has 1 of the 3 things, we will say that the text has no depression sintoms\n sentiment_model = pipeline(model=\"raulangelj/huggingface_sentiment_analysis\")\n is_negative_sentiment = sentiment_model(text)[0]['label'] == 'negative'\n # sentiment_model(data)\n is_depression_words = has_depression_words(text)\n is_first_person_pronouns = has_first_person_pronouns(text)\n # if all the conditions are true, then the text has depression sintoms\n if is_negative_sentiment and is_depression_words and is_first_person_pronouns:\n # return 'depression'\n return True\n # if 2 of the 3 conditions are true, then the text has possible depression sintoms\n # if (is_negative_sentiment and is_depression_words) or (is_negative_sentiment and is_first_person_pronouns) or (is_depression_words and is_first_person_pronouns):\n # # return 'possible_depression'\n # return True\n # if 1 of the 3 conditions are true, then the text has no depression sintoms\n if is_negative_sentiment or is_depression_words or is_first_person_pronouns:\n # return 'no_depression'\n return False\n # if none of the conditions are true, then the text has no depression sintoms\n # return 'no_depression'\n return False\n\n\n# -\n\nexample_text_depression = \"Últimamente, me cuesta encontrar motivación para levantarme de la cama por las mañanas. Siento una tristeza constante que no puedo sacudir, como si estuviera atrapado en una nube gris. Incluso las actividades que solían traerme alegría ahora me resultan abrumadoras y sin sentido. Me siento agotado todo el tiempo, tanto física como emocionalmente. La soledad se ha convertido en mi compañera constante, y me resulta difícil conectar con los demás. No puedo evitar sentir que he perdido el control de mi vida y que no hay esperanza de que las cosas mejoren. Estos pensamientos negativos y la sensación de vacío me invaden, y a veces me encuentro llorando sin razón aparente. Me preocupa que estos sentimientos persistan y no sé qué hacer para superar esta desesperanza abrumadora.\"\nexample_text_no_depression = \"El sol brilla radiante en el cielo azul mientras disfruto de un paseo por el parque. Río y converso con mis amigos, sintiéndome lleno de energía y entusiasmo. Me siento agradecido por las bendiciones en mi vida y tengo metas emocionantes que persigo con determinación. Cada día es una oportunidad para crecer, aprender y disfrutar de las pequeñas cosas que hacen que la vida sea maravillosa. Me rodeo de personas positivas que me apoyan y me animan en mis sueños. La vida es un viaje emocionante y estoy emocionado de ver lo que el futuro me depara.\"\nexample_text_no_depression_sad = \"Atravesé un momento difícil en mi vida recientemente. Experimenté una pérdida personal significativa que me dejó con el corazón apesadumbrado. Durante un tiempo, sentí una profunda tristeza y una sensación de vacío en mi interior. Sin embargo, me permití sentir y procesar estas emociones, encontrando consuelo en el amor y el apoyo de mis seres queridos. Poco a poco, comencé a encontrar la paz dentro de mí y a enfocarme en las cosas positivas que todavía tengo en mi vida. Aprendí a apreciar más cada momento y a encontrar la fortaleza para seguir adelante. Aunque todavía tengo momentos de tristeza, también encuentro alegría y esperanza en las pequeñas cosas de la vida. La adversidad me ha enseñado a ser más resiliente y a valorar cada experiencia, tanto positiva como negativa.\"\nprint(has_possible_depression_sintom(example_text_depression))\nprint(has_possible_depression_sintom(example_text_no_depression))\nprint(has_possible_depression_sintom(example_text_no_depression_sad))\n","repo_name":"raulangelj/analisis_sentimientos_depresion","sub_path":"sentimientos.ipynb","file_name":"sentimientos.ipynb","file_ext":"py","file_size_in_byte":12552,"program_lang":"python","lang":"en","doc_type":"code","stars":1,"dataset":"github-jupyter-script","pt":"4"}
+{"seq_id":"39042147056","text":"# + [markdown] id=\"view-in-github\" colab_type=\"text\"\n# \n\n# + colab={\"base_uri\": \"https://localhost:8080/\", \"height\": 1000} id=\"SkRplJmkb2XU\" outputId=\"77712dae-8557-48be-be78-5d19792c1c5a\"\nimport numpy as np\nimport tensorflow as tf\nfrom tensorflow import keras\nfrom tensorflow.keras import datasets, layers, models\nfrom tensorflow.keras.layers import Dense, Dropout, Activation, Flatten, Conv2D, MaxPooling2D\nimport matplotlib.pyplot as plt\nimport numpy as np\nfrom tensorflow.keras.preprocessing.image import ImageDataGenerator\n\nimg_width, img_height, img_num_channels = 32, 32, 3\n\n(x_train, y_train), (x_test, y_test) = datasets.cifar100.load_data()\n\nnum_classes =100\n\ny_train = keras.utils.to_categorical(y_train, num_classes)\ny_test = keras.utils.to_categorical(y_test, num_classes)\n\nx_train = x_train.astype('float32') / 255.\nx_test = x_test.astype('float32') / 255.\n\n\nmodel = tf.keras.models.Sequential() \nmodel.add(tf.keras.layers.Conv2D(32, (3, 3), padding='same', activation='relu', input_shape=input_shape)) \nmodel.add(Dropout(0.2)) \nmodel.add(tf.keras.layers.MaxPooling2D(pool_size=(2, 2), strides=(2,2))) \nmodel.add(tf.keras.layers.Conv2D(64, (3, 3), padding='same', activation='relu')) \nmodel.add(Dropout(0.2)) \nmodel.add(tf.keras.layers.MaxPooling2D(pool_size=(2, 2), strides=(2,2))) \nmodel.add(tf.keras.layers.Conv2D(128, (3, 3), padding='same', activation='relu')) \nmodel.add(Dropout(0.2)) \nmodel.add(tf.keras.layers.MaxPooling2D(pool_size=(2, 2), strides=(2,2)))\nmodel.add(tf.keras.layers.Conv2D(256, (3, 3), padding='same', activation='relu')) \nmodel.add(Dropout(0.2)) \nmodel.add(tf.keras.layers.MaxPooling2D(pool_size=(2, 2), strides=(2,2)))\nmodel.add(tf.keras.layers.Conv2D(512, (3, 3), padding='same', activation='relu')) \nmodel.add(Dropout(0.2)) \nmodel.add(tf.keras.layers.MaxPooling2D(pool_size=(2, 2), strides=(2,2)))\n\nmodel.add(tf.keras.layers.Flatten()) \nmodel.add(tf.keras.layers.Dense(128, activation=tf.nn.relu)) \nmodel.add(Dropout(0.2)) \nmodel.add(tf.keras.layers.Dense(256, activation=tf.nn.relu))\nmodel.add(Dropout(0.2)) \nmodel.add(tf.keras.layers.Dense(100, activation=tf.nn.softmax)) \nmodel.compile(loss='categorical_crossentropy', optimizer='adam', metrics=['accuracy']) \nmodel.build(input_shape=(None,32,32,1)) \n \nmodel.summary() \n\nfrom tensorflow.keras.preprocessing.image import ImageDataGenerator\ndatagen = ImageDataGenerator(\n featurewise_center=True,\n featurewise_std_normalization=True,\n rotation_range=20,\n width_shift_range=0.2,\n height_shift_range=0.2,\n horizontal_flip=True,\n validation_split=0.2)\n\ndatagen.fit(x_train)\n\nmodel.fit_generator(datagen.flow(x_train, y_train, batch_size=256),\n steps_per_epoch=x_train.shape[0] // 256,\n epochs=20,\n validation_data=(x_test, y_test)\n)\n# Generate generalization metrics\nscore = model.evaluate(x_test, y_test)\nprint('Test loss:', score[0]) \nprint('Test accuracy:', score[1])\n#print(f'Test loss: {score[0]} / Test accuracy: {score[1]}')\n\npredictions = model.predict([x_test])\n#print(predictions)\n\nprint(np.argmax(predictions[0]))\n\n\nimg_path = x_test[0]\nprint(img_path.shape)\nif(len(img_path.shape) == 3):\n plt.imshow(np.squeeze(img_path))\nelif(len(img_path.shape) == 2):\n plt.imshow(img_path)\nelse:\n print(\"Image cannot be shown\")\n","repo_name":"Yashwanth1102/CIFAR100-with-DataAugmentation","sub_path":"cifar100-with-dataaugmentation.ipynb","file_name":"cifar100-with-dataaugmentation.ipynb","file_ext":"py","file_size_in_byte":3559,"program_lang":"python","lang":"en","doc_type":"code","stars":0,"dataset":"github-jupyter-script","pt":"4"}
+{"seq_id":"22292961592","text":"#GENERAL Library\nimport pandas as pd\nimport numpy as np\nimport seaborn as sns\nimport matplotlib.pyplot as plt\nimport matplotlib as mpl\nimport random\nimport time\n\n# # OVERVIEW\n\n# #### Main Path\n\n#PATH PROCESS Library\nimport os\nimport os.path\nfrom pathlib import Path\nimport glob\nfrom scipy.io import loadmat\nimport nibabel as nib\nimport csv\n\nMain_Path = Path(\"C://Users//didel//Downloads//JUB University//3rd semster//ImageProcessing//SDOBenchmark-data-full\\SDOBenchmark-data-full//training\")\n\n# ##### Data CSV file Process\n\nMain_Train_CSV = list(Main_Path.glob(r\"**/*.csv\"))\n\nprint(len(Main_Train_CSV))\n\nTrain_CSV = Main_Train_CSV[0]\n\nprint(Train_CSV)\n\nReading_CSV = pd.read_csv(Train_CSV)\n\nprint(Reading_CSV.columns)\n\nprint(len(Reading_CSV))\n\nprint(Reading_CSV.head(-1))\n\nprint(Reading_CSV.isnull().sum())\n\nprint(Reading_CSV[\"peak_flux\"][0])\n\nprint(int(Reading_CSV[\"peak_flux\"][0]))\n\nprint(int(Reading_CSV[\"peak_flux\"][8334]))\n\nprint(Reading_CSV[\"start\"][8334])\n\nprint(Reading_CSV[\"start\"][8334][0:10])\n\n# * DATETIME TRANSFORMATION \n\nprint(type(Reading_CSV[\"start\"][8334][0:10]))\n\nTesting_Date_Transformation = Reading_CSV[\"start\"][8334][0:10].replace(\"-\",\"\")\n\nprint(Testing_Date_Transformation)\n\nTesting_Date_Array = np.array(int(Testing_Date_Transformation))\n\nprint(type(Testing_Date_Array))\n\n# #\n\n# * Data Process\n#\n# ###### Engineering\n\nJPG_Path = list(Main_Path.glob(r\"**/*.jpg\"))\n\nSorted_JPG_Path = sorted(JPG_Path)\n\nReading_CSV[\"New_ID\"] = Sorted_JPG_Path[:8336]\n\nprint(Reading_CSV.head(-1))\n\nprint(Reading_CSV[\"New_ID\"][0])\nprint(\"---\"*10)\nprint(Reading_CSV[\"id\"][0])\n\n# #\n\n# * New Data\n\n# +\nNew_Start = []\nNew_End = []\nNew_Path = []\n\nfor x_start,x_end,x_path in zip(Reading_CSV.start.values,Reading_CSV.end.values,Reading_CSV.New_ID.values):\n \n x_start = x_start[0:10]\n x_start = x_start.replace(\"-\",\"\")\n x_start = np.array(x_start,dtype=\"float32\")\n \n x_end = x_end[0:10]\n x_end = x_end.replace(\"-\",\"\")\n x_end = np.array(x_end,dtype=\"float32\")\n \n New_Start.append(x_start)\n New_End.append(x_end)\n New_Path.append(x_path)\n# -\n\nprint(\"LEN START: \",len(New_Start))\nprint(\"LEN END: \",len(New_End))\nprint(\"LEN PATH: \",len(New_Path))\n\nprint(New_Start[0])\nprint(\"--\"*10)\nprint(New_End[0])\nprint(\"--\"*10)\nprint(New_Path[0])\n\nprint(type(New_Start[0]))\nprint(\"--\"*10)\nprint(type(New_End[0]))\nprint(\"--\"*10)\nprint(type(New_Path[0]))\n\n# # \n\n# * To DataFrame\n\nStart_Series = pd.Series(New_Start,name=\"START\").astype(np.float32)\nEnd_Series = pd.Series(New_End,name=\"END\").astype(np.float32)\nPath_Series = pd.Series(New_Path,name=\"PATH\").astype(str)\n\nMain_Data = pd.concat([Path_Series,Start_Series,End_Series],axis=1)\n\nprint(Main_Data.head(-1))\n\n# # Vision\n\n# * Simple\n\n#IMAGE PROCESS Library\nfrom PIL import Image\nfrom keras.preprocessing import image\nfrom tensorflow.keras.preprocessing.image import ImageDataGenerator\nimport cv2\nfrom keras.applications.vgg16 import preprocess_input, decode_predictions\nfrom keras.preprocessing import image\nfrom skimage.feature import hessian_matrix, hessian_matrix_eigvals\nfrom scipy.ndimage.filters import convolve\nfrom skimage import data, io, filters\nimport skimage\nfrom skimage.morphology import convex_hull_image, erosion\nfrom IPython import display\nfrom scipy.ndimage import gaussian_filter\nfrom mpl_toolkits.mplot3d.art3d import Poly3DCollection\nimport matplotlib.patches as patches\n\n# +\nExample_IMG = cv2.cvtColor(cv2.imread(Main_Data[\"PATH\"][33]),cv2.COLOR_BGR2RGB)\n\nfigure = plt.figure(figsize=(10,10))\n\nplt.xlabel(Example_IMG.shape)\nplt.ylabel(Example_IMG.size)\nplt.imshow(Example_IMG)\n# -\n\n# * Multi\n\n# +\nfigure,axis = plt.subplots(5,5,figsize=(10,10))\n\nfor indexing,operations in enumerate(axis.flat):\n \n Reading_IMG = cv2.cvtColor(cv2.imread(Main_Data[\"PATH\"][indexing]),cv2.COLOR_BGR2RGB)\n \n operations.set_xlabel(Reading_IMG.shape)\n operations.set_ylabel(Reading_IMG.size)\n operations.imshow(Reading_IMG)\n\nplt.tight_layout()\nplt.show()\n# -\n\n# * 2D\n\n# #### How do scientists predict and track solar flares?\n# * Scientists typically track solar cycles by counting sunspots—flares of activity spurred by knots of magnetic field loops. The sunspot number climbs over the course of a solar cycle, then drops near zero as magnetic activity subsides.\n\n# +\nExample_IMG = cv2.cvtColor(cv2.imread(Main_Data[\"PATH\"][13]),cv2.COLOR_BGR2RGB)\n\nfigure = plt.figure(figsize=(10,10))\n\nplt.xlabel(Example_IMG[:,:,0].shape)\nplt.ylabel(Example_IMG[:,:,0].size)\nplt.imshow(Example_IMG[:,:,0])\n# -\n\n# * Threshold\n\n# +\nExample_IMG = cv2.cvtColor(cv2.imread(Main_Data[\"PATH\"][13]),cv2.COLOR_BGR2RGB)\n_,Threshold_IMG = cv2.threshold(Example_IMG,200,255,cv2.THRESH_BINARY)\n\nfigure = plt.figure(figsize=(10,10))\n\nplt.xlabel(Threshold_IMG.shape)\nplt.ylabel(Threshold_IMG.size)\nplt.imshow(Threshold_IMG)\n\n# +\nfigure,axis = plt.subplots(1,2,figsize=(10,10))\n\nExample_IMG = cv2.cvtColor(cv2.imread(Main_Data[\"PATH\"][13]),cv2.COLOR_BGR2RGB)\n_,Threshold_IMG = cv2.threshold(Example_IMG,200,255,cv2.THRESH_BINARY)\n\naxis[0].imshow(Example_IMG)\naxis[0].axis(\"off\")\naxis[1].imshow(Threshold_IMG)\naxis[1].axis(\"off\")\n\n# +\nfigure,axis = plt.subplots(1,2,figsize=(10,10))\n\nExample_IMG = cv2.cvtColor(cv2.imread(Main_Data[\"PATH\"][23]),cv2.COLOR_BGR2RGB)\n_,Threshold_IMG = cv2.threshold(Example_IMG,200,255,cv2.THRESH_BINARY)\n\naxis[0].imshow(Example_IMG)\naxis[0].axis(\"off\")\naxis[1].imshow(Threshold_IMG)\naxis[1].axis(\"off\")\n\n# +\nfigure,axis = plt.subplots(1,2,figsize=(10,10))\n\nExample_IMG = cv2.cvtColor(cv2.imread(Main_Data[\"PATH\"][2]),cv2.COLOR_BGR2RGB)\n_,Threshold_IMG = cv2.threshold(Example_IMG,200,255,cv2.THRESH_BINARY)\n\naxis[0].imshow(Example_IMG)\naxis[0].axis(\"off\")\naxis[1].imshow(Threshold_IMG)\naxis[1].axis(\"off\")\n\n# +\nfigure,axis = plt.subplots(1,2,figsize=(10,10))\n\nExample_IMG = cv2.cvtColor(cv2.imread(Main_Data[\"PATH\"][404]),cv2.COLOR_BGR2RGB)\n_,Threshold_IMG = cv2.threshold(Example_IMG,200,255,cv2.THRESH_BINARY)\n\naxis[0].imshow(Example_IMG)\naxis[0].axis(\"off\")\naxis[1].imshow(Threshold_IMG)\naxis[1].axis(\"off\")\n# -\n\n# * Hessian Spectrum \n\n# +\nfigure,axis = plt.subplots(1,3,figsize=(10,10))\n\nExample_IMG = cv2.cvtColor(cv2.imread(Main_Data[\"PATH\"][404]),cv2.COLOR_BGR2GRAY)\nHessian_IMG = hessian_matrix(Example_IMG,sigma=0.20,order=\"rc\")\nMax_IMG,Min_IMG = hessian_matrix_eigvals(Hessian_IMG)\n\naxis[0].imshow(Example_IMG)\naxis[0].axis(\"off\")\naxis[1].imshow(Max_IMG)\naxis[1].axis(\"off\")\naxis[2].imshow(Min_IMG)\naxis[2].axis(\"off\")\n\n# +\nfigure,axis = plt.subplots(1,3,figsize=(10,10))\n\nExample_IMG = cv2.cvtColor(cv2.imread(Main_Data[\"PATH\"][44]),cv2.COLOR_BGR2GRAY)\nHessian_IMG = hessian_matrix(Example_IMG,sigma=0.20,order=\"rc\")\nMax_IMG,Min_IMG = hessian_matrix_eigvals(Hessian_IMG)\n\naxis[0].imshow(Example_IMG)\naxis[0].axis(\"off\")\naxis[1].imshow(Max_IMG)\naxis[1].axis(\"off\")\naxis[2].imshow(Min_IMG)\naxis[2].axis(\"off\")\n# -\n\n# * Canny-Threshold Concat\n\n# +\nfigure,axis = plt.subplots(1,3,figsize=(10,10))\n\nExample_IMG = cv2.cvtColor(cv2.imread(Main_Data[\"PATH\"][44]),cv2.COLOR_BGR2RGB)\n_,Threshold_IMG = cv2.threshold(Example_IMG,200,255,cv2.THRESH_BINARY)\nCanny_IMG = cv2.Canny(Threshold_IMG,10,100)\n\n\naxis[0].imshow(Example_IMG)\naxis[0].axis(\"off\")\naxis[1].imshow(Threshold_IMG)\naxis[1].axis(\"off\")\naxis[2].imshow(Canny_IMG)\naxis[2].axis(\"off\")\n\n# +\nfigure,axis = plt.subplots(1,3,figsize=(10,10))\n\nExample_IMG = cv2.cvtColor(cv2.imread(Main_Data[\"PATH\"][333]),cv2.COLOR_BGR2RGB)\n_,Threshold_IMG = cv2.threshold(Example_IMG,200,255,cv2.THRESH_BINARY)\nCanny_IMG = cv2.Canny(Threshold_IMG,10,100)\n\n\naxis[0].imshow(Example_IMG)\naxis[0].axis(\"off\")\naxis[1].imshow(Threshold_IMG)\naxis[1].axis(\"off\")\naxis[2].imshow(Canny_IMG)\naxis[2].axis(\"off\")\n\n# +\nfigure,axis = plt.subplots(1,3,figsize=(15,15))\n\nExample_IMG = cv2.cvtColor(cv2.imread(Main_Data[\"PATH\"][333]),cv2.COLOR_BGR2RGB)\n_,Threshold_IMG = cv2.threshold(Example_IMG,200,255,cv2.THRESH_BINARY)\nCanny_IMG = cv2.Canny(Threshold_IMG,10,100)\nConcat_IMG = cv2.addWeighted(Example_IMG[:,:,0],0.8,Canny_IMG,0.4,0.5)\n\naxis[0].imshow(Example_IMG)\naxis[0].axis(\"off\")\naxis[1].imshow(Canny_IMG)\naxis[1].axis(\"off\")\naxis[2].imshow(Concat_IMG)\naxis[2].axis(\"off\")\n# -\n\n# * Contours\n\n# +\nfigure,axis = plt.subplots(1,3,figsize=(15,15))\n\nExample_IMG = cv2.cvtColor(cv2.imread(Main_Data[\"PATH\"][13]),cv2.COLOR_BGR2RGB)\n_,Threshold_IMG = cv2.threshold(Example_IMG,200,255,cv2.THRESH_BINARY)\nCanny_IMG = cv2.Canny(Threshold_IMG,10,100)\nConcat_IMG = cv2.addWeighted(Example_IMG[:,:,0],0.8,Canny_IMG,0.4,0.5)\ncontours,_ = cv2.findContours(Canny_IMG,cv2.RETR_EXTERNAL,cv2.CHAIN_APPROX_SIMPLE)\n\nfor contour in contours:\n x,y,w,h = cv2.boundingRect(contour)\n Rect_IMG = cv2.rectangle(Concat_IMG,(x,y),(x+w,y+h),(255,0,0),2)\n \naxis[0].imshow(Example_IMG)\naxis[0].axis(\"off\")\naxis[1].imshow(Concat_IMG,cmap=\"hot\")\naxis[1].axis(\"off\")\naxis[2].imshow(Threshold_IMG)\naxis[2].axis(\"off\")\n\n# +\nfigure,axis = plt.subplots(1,3,figsize=(15,15))\n\nExample_IMG = cv2.cvtColor(cv2.imread(Main_Data[\"PATH\"][333]),cv2.COLOR_BGR2RGB)\n_,Threshold_IMG = cv2.threshold(Example_IMG,200,255,cv2.THRESH_BINARY)\nCanny_IMG = cv2.Canny(Threshold_IMG,10,100)\nConcat_IMG = cv2.addWeighted(Example_IMG[:,:,0],0.8,Canny_IMG,0.4,0.5)\ncontours,_ = cv2.findContours(Canny_IMG,cv2.RETR_EXTERNAL,cv2.CHAIN_APPROX_SIMPLE)\n\nfor contour in contours:\n x,y,w,h = cv2.boundingRect(contour)\n Rect_IMG = cv2.rectangle(Concat_IMG,(x,y),(x+w,y+h),(255,0,0),2)\n \naxis[0].imshow(Example_IMG)\naxis[0].axis(\"off\")\naxis[1].imshow(Concat_IMG,cmap=\"hot\")\naxis[1].axis(\"off\")\naxis[2].imshow(Threshold_IMG)\naxis[2].axis(\"off\")\n\n# +\nfigure,axis = plt.subplots(1,3,figsize=(15,15))\n\nExample_IMG = cv2.cvtColor(cv2.imread(Main_Data[\"PATH\"][113]),cv2.COLOR_BGR2RGB)\n_,Threshold_IMG = cv2.threshold(Example_IMG,200,255,cv2.THRESH_BINARY)\nCanny_IMG = cv2.Canny(Threshold_IMG,10,100)\nConcat_IMG = cv2.addWeighted(Example_IMG[:,:,0],0.8,Canny_IMG,0.4,0.5)\ncontours,_ = cv2.findContours(Canny_IMG,cv2.RETR_EXTERNAL,cv2.CHAIN_APPROX_SIMPLE)\n\nfor contour in contours:\n x,y,w,h = cv2.boundingRect(contour)\n Rect_IMG = cv2.rectangle(Concat_IMG,(x,y),(x+w,y+h),(255,0,0),2)\n \naxis[0].imshow(Example_IMG)\naxis[0].axis(\"off\")\naxis[1].imshow(Concat_IMG,cmap=\"hot\")\naxis[1].axis(\"off\")\naxis[2].imshow(Threshold_IMG)\naxis[2].axis(\"off\")\n\n# +\nfigure,axis = plt.subplots(1,3,figsize=(15,15))\n\nExample_IMG = cv2.cvtColor(cv2.imread(Main_Data[\"PATH\"][188]),cv2.COLOR_BGR2RGB)\n_,Threshold_IMG = cv2.threshold(Example_IMG,200,255,cv2.THRESH_BINARY)\nCanny_IMG = cv2.Canny(Threshold_IMG,10,100)\nConcat_IMG = cv2.addWeighted(Example_IMG[:,:,0],0.8,Canny_IMG,0.4,0.5)\ncontours,_ = cv2.findContours(Canny_IMG,cv2.RETR_EXTERNAL,cv2.CHAIN_APPROX_SIMPLE)\n\nfor contour in contours:\n x,y,w,h = cv2.boundingRect(contour)\n Rect_IMG = cv2.rectangle(Concat_IMG,(x,y),(x+w,y+h),(255,0,0),2)\n \naxis[0].imshow(Example_IMG)\naxis[0].axis(\"off\")\naxis[1].imshow(Concat_IMG,cmap=\"hot\")\naxis[1].axis(\"off\")\naxis[2].imshow(Threshold_IMG)\naxis[2].axis(\"off\")\n# -\n\n# * Layer Concat\n\n# +\nfigure,axis = plt.subplots(1,3,figsize=(15,15))\n\nExample_IMG = cv2.cvtColor(cv2.imread(Main_Data[\"PATH\"][13]),cv2.COLOR_BGR2RGB)\n_,Threshold_IMG = cv2.threshold(Example_IMG,200,255,cv2.THRESH_BINARY)\nCanny_IMG = cv2.Canny(Threshold_IMG,10,100)\n\nCopy_Main = Example_IMG.copy()\nCopy_Main[Canny_IMG == 1] = [0,255,0]\nCopy_Main[Canny_IMG == 2] = [0,0,255]\nCopy_Main_Two = Example_IMG.copy()\n\nConcat_IMG = cv2.addWeighted(Copy_Main,0.5,Copy_Main_Two,0.7,0,Copy_Main_Two)\n\naxis[0].imshow(Example_IMG)\naxis[0].axis(\"off\")\naxis[1].imshow(Canny_IMG)\naxis[1].axis(\"off\")\naxis[2].imshow(Concat_IMG)\naxis[2].axis(\"off\")\n\n# +\nfigure,axis = plt.subplots(1,3,figsize=(15,15))\n\nExample_IMG = cv2.cvtColor(cv2.imread(Main_Data[\"PATH\"][13]),cv2.COLOR_BGR2RGB)\n_,Threshold_IMG = cv2.threshold(Example_IMG,200,255,cv2.THRESH_BINARY)\nCanny_IMG = cv2.Canny(Threshold_IMG,10,100)\n\nCopy_Main = Example_IMG.copy()\nCopy_Main[Canny_IMG == 255] = [255,0,0]\n\naxis[0].imshow(Example_IMG)\naxis[0].axis(\"off\")\naxis[1].imshow(Canny_IMG)\naxis[1].axis(\"off\")\naxis[2].imshow(Copy_Main)\naxis[2].axis(\"off\")\n\n# +\nfigure,axis = plt.subplots(1,3,figsize=(15,15))\n\nExample_IMG = cv2.cvtColor(cv2.imread(Main_Data[\"PATH\"][123]),cv2.COLOR_BGR2RGB)\n_,Threshold_IMG = cv2.threshold(Example_IMG,200,255,cv2.THRESH_BINARY)\nCanny_IMG = cv2.Canny(Threshold_IMG,10,100)\n\nCopy_Main = Example_IMG.copy()\nCopy_Main[Canny_IMG == 255] = [255,0,0]\n\naxis[0].imshow(Example_IMG)\naxis[0].axis(\"off\")\naxis[1].imshow(Canny_IMG)\naxis[1].axis(\"off\")\naxis[2].imshow(Copy_Main)\naxis[2].axis(\"off\")\n\n# +\nfigure,axis = plt.subplots(1,3,figsize=(15,15))\n\nExample_IMG = cv2.cvtColor(cv2.imread(Main_Data[\"PATH\"][223]),cv2.COLOR_BGR2RGB)\n_,Threshold_IMG = cv2.threshold(Example_IMG,200,255,cv2.THRESH_BINARY)\nCanny_IMG = cv2.Canny(Threshold_IMG,10,100)\n\nCopy_Main = Example_IMG.copy()\nCopy_Main[Threshold_IMG[:,:,0] == 255] = [255,0,0]\n\naxis[0].imshow(Example_IMG)\naxis[0].axis(\"off\")\naxis[1].imshow(Threshold_IMG)\naxis[1].axis(\"off\")\naxis[2].imshow(Copy_Main)\naxis[2].axis(\"off\")\n\n# +\nfigure,axis = plt.subplots(1,3,figsize=(15,15))\n\nExample_IMG = cv2.cvtColor(cv2.imread(Main_Data[\"PATH\"][1]),cv2.COLOR_BGR2RGB)\n_,Threshold_IMG = cv2.threshold(Example_IMG,200,255,cv2.THRESH_BINARY)\nCanny_IMG = cv2.Canny(Threshold_IMG,10,100)\n\nCopy_Main = Example_IMG.copy()\nCopy_Main[Threshold_IMG[:,:,0] == 255] = [255,0,0]\n\naxis[0].imshow(Example_IMG)\naxis[0].axis(\"off\")\naxis[1].imshow(Threshold_IMG)\naxis[1].axis(\"off\")\naxis[2].imshow(Copy_Main)\naxis[2].axis(\"off\")\n\n# +\nfigure,axis = plt.subplots(1,3,figsize=(15,15))\n\nExample_IMG = cv2.cvtColor(cv2.imread(Main_Data[\"PATH\"][17]),cv2.COLOR_BGR2RGB)\n_,Threshold_IMG = cv2.threshold(Example_IMG,200,255,cv2.THRESH_BINARY)\nCanny_IMG = cv2.Canny(Threshold_IMG,10,100)\n\nCopy_Main = Example_IMG.copy()\nCopy_Main[Threshold_IMG[:,:,0] == 255] = [255,0,0]\n\naxis[0].imshow(Example_IMG)\naxis[0].axis(\"off\")\naxis[1].imshow(Threshold_IMG)\naxis[1].axis(\"off\")\naxis[2].imshow(Copy_Main)\naxis[2].axis(\"off\")\n# -\n\n# # Data Transformation\n\n# * For Prediction\n\n#SCALER & TRANSFORMATION Library\nfrom sklearn.preprocessing import StandardScaler\nfrom sklearn.preprocessing import MinMaxScaler\nfrom keras.utils.np_utils import to_categorical\nfrom sklearn.model_selection import train_test_split\nfrom keras import regularizers\nfrom sklearn.preprocessing import LabelEncoder\n\n# +\nX_Start = []\nX_End = []\nX_Image = []\n\nfor img,start_i,end_i in zip(Main_Data.PATH.values,Main_Data.START.values,Main_Data.END.values):\n Picking_IMG = cv2.cvtColor(cv2.imread(img),cv2.COLOR_BGR2RGB)\n Picking_IMG = cv2.resize(Picking_IMG,(180,180))\n Picking_IMG = Picking_IMG / 255.\n \n X_Image.append(Picking_IMG)\n X_Start.append(start_i)\n X_End.append(end_i)\n# -\n\nprint(np.shape(np.array(X_Image)))\nprint(np.shape(np.array(X_Start)))\nprint(np.shape(np.array(X_End)))\n\nTrain_JPG = np.array(X_Image,dtype=\"float32\")\nTrain_START = np.array(X_Start,dtype=\"float32\")\nTrain_END = np.array(X_End,dtype=\"float32\")\n\nprint(Train_JPG.shape)\nprint(Train_START.shape)\nprint(Train_END.shape)\n\n# +\nScalar_Function = MinMaxScaler()\n\nTrain_START_R = Scalar_Function.fit_transform(Train_START.reshape(-1,1))\nTrain_END_R = Scalar_Function.fit_transform(Train_END.reshape(-1,1))\n# -\n\n# * For Auto-Encoder\n\n# +\nX_Mask = []\nX_New_IMG = []\n\nfor img_i in Main_Data.PATH.values:\n Picking_IMG = cv2.cvtColor(cv2.imread(img_i),cv2.COLOR_BGR2RGB)\n _,Threshold_IMG = cv2.threshold(Picking_IMG,200,255,cv2.THRESH_BINARY)\n Canny_IMG = cv2.Canny(Threshold_IMG,10,100)\n\n Copy_Main = Picking_IMG.copy()\n Copy_Main[Canny_IMG == 255] = [255,0,0]\n \n Copy_Main = cv2.resize(Copy_Main,(180,180))\n Copy_Main = Copy_Main / 255.\n \n Picking_IMG = cv2.resize(Picking_IMG,(180,180))\n Picking_IMG = Picking_IMG / 255.\n \n X_Mask.append(Copy_Main)\n X_New_IMG.append(Picking_IMG)\n\n# +\nX_Mask = []\nX_New_IMG = []\n\nbatch_size = 1 \nfor start_idx in range(0, len(Main_Data.PATH.values), batch_size):\n end_idx = start_idx + batch_size\n \n # Process images in the current batch\n for img_path in Main_Data.PATH.values[start_idx:end_idx]:\n Picking_IMG = cv2.cvtColor(cv2.imread(img_path), cv2.COLOR_BGR2RGB)\n _, Threshold_IMG = cv2.threshold(Picking_IMG, 200, 255, cv2.THRESH_BINARY)\n Canny_IMG = cv2.Canny(Threshold_IMG, 10, 100)\n\n Copy_Main = Picking_IMG.copy()\n Copy_Main[Canny_IMG == 255] = [255, 0, 0]\n\n Copy_Main = cv2.resize(Copy_Main, (180, 180))\n Copy_Main = Copy_Main / 255.\n\n Picking_IMG = cv2.resize(Picking_IMG, (180, 180))\n Picking_IMG = Picking_IMG / 255.\n\n X_Mask.append(Copy_Main)\n X_New_IMG.append(Picking_IMG)\n\n# After the loop, X_Mask and X_New_IMG will contain the processed images\n\n# -\n\nX_Mask = np.array(X_Mask,dtype=\"float32\")\nX_New_IMG = np.array(X_New_IMG,dtype=\"float32\")\n\n\n\n\n\n#ACCURACY CONTROL library\nfrom sklearn.metrics import confusion_matrix, accuracy_score, classification_report, roc_auc_score, roc_curve\nfrom sklearn.model_selection import GridSearchCV, cross_val_score\nfrom sklearn.metrics import mean_squared_error, r2_score\n\n#OPTIMIZER library\nfrom keras.optimizers import RMSprop,Adam,Optimizer,Optimizer, SGD\n\n#MODEL LAYERS Library\nfrom tensorflow.keras.models import Sequential\nfrom keras.layers import Dense, Dropout, Flatten, Conv2D, MaxPool2D, BatchNormalization,MaxPooling2D,BatchNormalization,\\\n Permute, TimeDistributed, Bidirectional,GRU, SimpleRNN,\\\nLSTM, GlobalAveragePooling2D, SeparableConv2D, ZeroPadding2D, Convolution2D, ZeroPadding2D,Reshape, Conv2DTranspose,\\\nLeakyReLU, GaussianNoise, GlobalMaxPooling2D, ReLU, Input, Concatenate\nfrom keras import models\nfrom keras import layers\nimport tensorflow as tf\nfrom keras.applications import VGG16,VGG19,inception_v3\nfrom keras import backend as K\nfrom keras.utils import plot_model\nfrom keras.datasets import mnist\nimport keras\nfrom keras.models import Model\n\n#IGNORING WARNINGS Library\nfrom warnings import filterwarnings\nfilterwarnings(\"ignore\",category=DeprecationWarning)\nfilterwarnings(\"ignore\", category=FutureWarning) \nfilterwarnings(\"ignore\", category=UserWarning)\n","repo_name":"radidel/radidel.github.io","sub_path":"Solar_flares.ipynb","file_name":"Solar_flares.ipynb","file_ext":"py","file_size_in_byte":18055,"program_lang":"python","lang":"en","doc_type":"code","stars":0,"dataset":"github-jupyter-script","pt":"40"}
+{"seq_id":"14621573869","text":"# + pycharm={\"name\": \"#%%\\n\"}\n# Import necessary libraries and modules\nimport matplotlib.pyplot as plt\nimport networkx as nx\nimport numpy as np\nimport pandas as pd\nimport tadasets\nfrom sklearn.metrics import pairwise_distances\n\n\n# + pycharm={\"name\": \"#%%\\n\"}\n# Function to create a Vietoris-Rips complex given a distance matrix and radius\ndef create_simplicial_complex(D, r):\n \"\"\"\n Input: distance matrix and not negative radius\n Output: networkx graph\n \"\"\"\n\n G = nx.Graph()\n G.add_nodes_from(list(range(len(D))))\n edge_list = np.argwhere(D <= r)\n G.add_edges_from(edge_list)\n\n # Remove self-loops\n G.remove_edges_from(nx.selfloop_edges(G))\n\n return G\n\n\n# + pycharm={\"name\": \"#%%\\n\"}\n# Generate three sets of 2D points with different means and covariances\nX1 = np.random.multivariate_normal([0, 0], np.array([[1, 0], [0, 1]]), size=10)\nX2 = np.random.multivariate_normal([7, 0], np.array([[1, 0], [0, 0.2]]), size=8)\nX3 = np.random.multivariate_normal([0, 11], np.array([[1, 0], [0, 1]]), size=12)\n\n# Combine the three sets of points into one array\nX = np.concatenate((X1, X2, X3))\n\n# Plot the combined points\nplt.plot(X[:, 0], X[:, 1], 'o')\nplt.axis('equal')\nplt.show()\n\n# + pycharm={\"name\": \"#%%\\n\"}\nX\n\n# + pycharm={\"name\": \"#%%\\n\"}\n# Set random seed for reproducibility\nnp.random.seed(9)\n\n# Generate points from a noisy d-sphere using Tadasets library\nx = tadasets.dsphere(n=50, d=1, noise=0.1)\ny = np.array([[1, 0], [0, 1], [2, 2]])\n\n# Plot the points from the d-sphere\nplt.scatter(x[:, 0], x[:, 1])\nplt.axis('equal')\nplt.show()\n\n# + pycharm={\"name\": \"#%%\\n\"}\nnuevox = pd.DataFrame(x)\nnuevoy = pd.DataFrame(y)\n\n# + pycharm={\"name\": \"#%%\\n\"}\nnuevox\n\n# + pycharm={\"name\": \"#%%\\n\"}\nnuevoy\n\n# + pycharm={\"name\": \"#%%\\n\"}\n# Convert x and y (another dataset) to pandas dataframes\nnuevox = pd.DataFrame(x)\nnuevoy = pd.DataFrame(y)\n\n# Calculate pairwise distance matrices for x and y\ndist = pairwise_distances(x)\nplt.imshow(dist)\nplt.show()\n\n# + pycharm={\"name\": \"#%%\\n\"}\ny = pairwise_distances(y)\nplt.imshow(y)\nplt.show()\n\n# + pycharm={\"name\": \"#%%\\n\"}\n# Create and plot Vietoris-Rips complex for radius 0.5\nr = 0.5\nG = create_simplicial_complex(dist, r)\nnx.draw_kamada_kawai(G)\n\n# + pycharm={\"name\": \"#%%\\n\"}\nposi = {n: x[n, :] for n in range(len(x))}\nplt.figure(figsize=(5, 5))\n\nnx.draw_networkx(G, pos=posi, with_labels=False, node_size=20)\nplt.axis('equal')\nplt.show()\n\n# + pycharm={\"name\": \"#%%\\n\"}\n# Create and plot Vietoris-Rips complex for radius 0.7\nr = 0.7\nG = create_simplicial_complex(dist, r)\nnx.draw_kamada_kawai(G)\n\n# + pycharm={\"name\": \"#%%\\n\"}\nposi = {n: x[n, :] for n in range(len(x))}\nplt.figure(figsize=(5, 5))\n\nnx.draw_networkx(G, pos=posi, with_labels=False, node_size=20)\nplt.axis('equal')\nplt.show()\n\n# + pycharm={\"name\": \"#%%\\n\"}\n# Create and plot Vietoris-Rips complex for radius 1\nr = 1\nG = create_simplicial_complex(dist, r)\nnx.draw_kamada_kawai(G)\n\n# + pycharm={\"name\": \"#%%\\n\"}\nposi = {n: x[n, :] for n in range(len(x))}\nplt.figure(figsize=(5, 5))\n\nnx.draw_networkx(G, pos=posi, with_labels=False, node_size=20)\nplt.axis('equal')\nplt.show()\n","repo_name":"cesar-xyz/Topologia","sub_path":"complejoSimplicial.ipynb","file_name":"complejoSimplicial.ipynb","file_ext":"py","file_size_in_byte":3101,"program_lang":"python","lang":"en","doc_type":"code","stars":0,"dataset":"github-jupyter-script","pt":"40"}
+{"seq_id":"7754447719","text":"# + id=\"jjqL4lTWDUhP\" colab_type=\"code\" outputId=\"16a84d47-f240-41cc-8b3b-b7310aa532fb\" executionInfo={\"status\": \"ok\", \"timestamp\": 1587542145260, \"user_tz\": -480, \"elapsed\": 3244, \"user\": {\"displayName\": \"Liao Jack\", \"photoUrl\": \"\", \"userId\": \"16157886839679822522\"}} colab={\"base_uri\": \"https://localhost:8080/\", \"height\": 50}\nimport keras\nkeras.__version__\n\n# + [markdown] id=\"1fFvz2ojDUhU\" colab_type=\"text\"\n# # Sequence processing with convnets\n# ## Implementing a 1D convnet\n#\n# In Keras, you would use a 1D convnet via the `Conv1D` layer, which has a very similar interface to `Conv2D`. It takes as input 3D tensors \n# with shape `(samples, time, features)` and also returns similarly-shaped 3D tensors. The convolution window is a 1D window on the temporal \n# axis, axis 1 in the input tensor.\n#\n# Let's build a simple 2-layer 1D convnet and apply it to the IMDB sentiment classification task that you are already familiar with.\n#\n# As a reminder, this is the code for obtaining and preprocessing the data:\n\n# + id=\"GeEdmRH7DUhU\" colab_type=\"code\" outputId=\"5e33ed47-bc4a-4d1c-9c80-b42b72160ff7\" executionInfo={\"status\": \"ok\", \"timestamp\": 1587542154704, \"user_tz\": -480, \"elapsed\": 12675, \"user\": {\"displayName\": \"Liao Jack\", \"photoUrl\": \"\", \"userId\": \"16157886839679822522\"}} colab={\"base_uri\": \"https://localhost:8080/\", \"height\": 151}\nfrom keras.datasets import imdb\nfrom keras.preprocessing import sequence\n\nmax_features = 10000 # number of words to consider as features\nmax_len = 500 # cut texts after this number of words (among top max_features most common words)\n\nprint('Loading data...')\n(x_train, y_train), (x_test, y_test) = imdb.load_data(num_words=max_features)\nprint(len(x_train), 'train sequences')\nprint(len(x_test), 'test sequences')\n\nprint('Pad sequences (samples x time)')\nx_train = sequence.pad_sequences(x_train, maxlen=max_len)\nx_test = sequence.pad_sequences(x_test, maxlen=max_len)\nprint('x_train shape:', x_train.shape)\nprint('x_test shape:', x_test.shape)\n\n# + [markdown] id=\"dBYaA5zPDUhY\" colab_type=\"text\"\n# This is our example 1D convnet for the IMDB dataset:\n\n# + id=\"9MvCtNJeDUhY\" colab_type=\"code\" outputId=\"d8477977-bbbd-4268-bfd0-f33b38ffa327\" executionInfo={\"status\": \"ok\", \"timestamp\": 1587542198373, \"user_tz\": -480, \"elapsed\": 56334, \"user\": {\"displayName\": \"Liao Jack\", \"photoUrl\": \"\", \"userId\": \"16157886839679822522\"}} colab={\"base_uri\": \"https://localhost:8080/\", \"height\": 759}\nfrom keras.models import Sequential\nfrom keras import layers\nfrom keras.optimizers import RMSprop\n\nmodel = Sequential()\nmodel.add(layers.Embedding(max_features, 128, input_length=max_len))\nmodel.add(layers.Conv1D(32, 7, activation='relu'))\nmodel.add(layers.MaxPooling1D(5))\nmodel.add(layers.Conv1D(32, 7, activation='relu'))\nmodel.add(layers.GlobalMaxPooling1D())\nmodel.add(layers.Dense(1))\n\nmodel.summary()\n\nmodel.compile(optimizer=RMSprop(lr=1e-4),\n loss='binary_crossentropy',\n metrics=['acc'])\nhistory = model.fit(x_train, y_train,\n epochs=10,\n batch_size=128,\n validation_split=0.2)\n\n# + [markdown] id=\"aoOijqtYDUhb\" colab_type=\"text\"\n# 繪圖\n\n# + id=\"iyXKwEU4DUhb\" colab_type=\"code\" outputId=\"b54fb5b1-a7d5-432d-8f17-f6b4f566f810\" executionInfo={\"status\": \"ok\", \"timestamp\": 1587542198817, \"user_tz\": -480, \"elapsed\": 56767, \"user\": {\"displayName\": \"Liao Jack\", \"photoUrl\": \"\", \"userId\": \"16157886839679822522\"}} colab={\"base_uri\": \"https://localhost:8080/\", \"height\": 545}\nimport matplotlib.pyplot as plt\n\nacc = history.history['acc']\nval_acc = history.history['val_acc']\nloss = history.history['loss']\nval_loss = history.history['val_loss']\n\nepochs = range(len(acc))\n\nplt.plot(epochs, acc, 'bo', label='Training acc')\nplt.plot(epochs, val_acc, 'b', label='Validation acc')\nplt.title('Training and validation accuracy')\nplt.legend()\n\nplt.figure()\n\nplt.plot(epochs, loss, 'bo', label='Training loss')\nplt.plot(epochs, val_loss, 'b', label='Validation loss')\nplt.title('Training and validation loss')\nplt.legend()\n\nplt.show()\n\n# + [markdown] id=\"sma33T51DUhf\" colab_type=\"text\"\n# ## Combining CNNs and RNNs to process long sequences\n\n# + id=\"bA7PgyAXkn1Z\" colab_type=\"code\" outputId=\"96057e29-e4c4-4c8a-b998-79f7b97a98b8\" executionInfo={\"status\": \"ok\", \"timestamp\": 1587542204980, \"user_tz\": -480, \"elapsed\": 62917, \"user\": {\"displayName\": \"Liao Jack\", \"photoUrl\": \"\", \"userId\": \"16157886839679822522\"}} colab={\"base_uri\": \"https://localhost:8080/\", \"height\": 235}\n# !wget https://s3.amazonaws.com/keras-datasets/jena_climate_2009_2016.csv.zip\n# !unzip jena_climate_2009_2016.csv.zip\n\n# + id=\"--7RfsIpDUhg\" colab_type=\"code\" colab={}\n# We reuse the following variables defined in the last section:\n# float_data, train_gen, val_gen, val_steps\n\nimport os\nimport numpy as np\n\ndata_dir = './'\nfname = os.path.join(data_dir, 'jena_climate_2009_2016.csv')\n\nf = open(fname)\ndata = f.read()\nf.close()\n\nlines = data.split('\\n')\nheader = lines[0].split(',')\nlines = lines[1:]\n\nfloat_data = np.zeros((len(lines), len(header) - 1))\nfor i, line in enumerate(lines):\n values = [float(x) for x in line.split(',')[1:]]\n float_data[i, :] = values\n \nmean = float_data[:200000].mean(axis=0)\nfloat_data -= mean\nstd = float_data[:200000].std(axis=0)\nfloat_data /= std\n\ndef generator(data, lookback, delay, min_index, max_index,\n shuffle=False, batch_size=128, step=6):\n if max_index is None:\n max_index = len(data) - delay - 1\n i = min_index + lookback\n while 1:\n if shuffle:\n rows = np.random.randint(\n min_index + lookback, max_index, size=batch_size)\n else:\n if i + batch_size >= max_index:\n i = min_index + lookback\n rows = np.arange(i, min(i + batch_size, max_index))\n i += len(rows)\n\n samples = np.zeros((len(rows),\n lookback // step,\n data.shape[-1]))\n targets = np.zeros((len(rows),))\n for j, row in enumerate(rows):\n indices = range(rows[j] - lookback, rows[j], step)\n samples[j] = data[indices]\n targets[j] = data[rows[j] + delay][1]\n yield samples, targets\n \nlookback = 1440\nstep = 6\ndelay = 144\nbatch_size = 128\n\ntrain_gen = generator(float_data,\n lookback=lookback,\n delay=delay,\n min_index=0,\n max_index=200000,\n shuffle=True,\n step=step, \n batch_size=batch_size)\nval_gen = generator(float_data,\n lookback=lookback,\n delay=delay,\n min_index=200001,\n max_index=300000,\n step=step,\n batch_size=batch_size)\ntest_gen = generator(float_data,\n lookback=lookback,\n delay=delay,\n min_index=300001,\n max_index=None,\n step=step,\n batch_size=batch_size)\n\n# This is how many steps to draw from `val_gen`\n# in order to see the whole validation set:\nval_steps = (300000 - 200001 - lookback) // batch_size\n\n# This is how many steps to draw from `test_gen`\n# in order to see the whole test set:\ntest_steps = (len(float_data) - 300001 - lookback) // batch_size\n\n# + id=\"pPJQsaN7DUhj\" colab_type=\"code\" outputId=\"d6ddcb57-9fd9-4dd2-c00f-2c9ba6c6c494\" executionInfo={\"status\": \"ok\", \"timestamp\": 1587542495834, \"user_tz\": -480, \"elapsed\": 353759, \"user\": {\"displayName\": \"Liao Jack\", \"photoUrl\": \"\", \"userId\": \"16157886839679822522\"}} colab={\"base_uri\": \"https://localhost:8080/\", \"height\": 689}\nfrom keras.models import Sequential\nfrom keras import layers\nfrom keras.optimizers import RMSprop\n\nmodel = Sequential()\nmodel.add(layers.Conv1D(32, 5, activation='relu',\n input_shape=(None, float_data.shape[-1])))\nmodel.add(layers.MaxPooling1D(3))\nmodel.add(layers.Conv1D(32, 5, activation='relu'))\nmodel.add(layers.MaxPooling1D(3))\nmodel.add(layers.Conv1D(32, 5, activation='relu'))\nmodel.add(layers.GlobalMaxPooling1D())\nmodel.add(layers.Dense(1))\n\nmodel.compile(optimizer=RMSprop(), loss='mae')\nhistory = model.fit_generator(train_gen,\n steps_per_epoch=500,\n epochs=20,\n validation_data=val_gen,\n validation_steps=val_steps)\n\n# + [markdown] id=\"r54fhB_cDUho\" colab_type=\"text\"\n# Here are our training and validation Mean Absolute Errors:\n\n# + id=\"r7yK_n2pDUhp\" colab_type=\"code\" outputId=\"edbf76e5-f9a9-4c6d-8ea5-38f729829bce\" executionInfo={\"status\": \"ok\", \"timestamp\": 1587542495837, \"user_tz\": -480, \"elapsed\": 353753, \"user\": {\"displayName\": \"Liao Jack\", \"photoUrl\": \"\", \"userId\": \"16157886839679822522\"}} colab={\"base_uri\": \"https://localhost:8080/\", \"height\": 281}\nimport matplotlib.pyplot as plt\n\nloss = history.history['loss']\nval_loss = history.history['val_loss']\n\nepochs = range(len(loss))\n\nplt.figure()\n\nplt.plot(epochs, loss, 'bo', label='Training loss')\nplt.plot(epochs, val_loss, 'b', label='Validation loss')\nplt.title('Training and validation loss')\nplt.legend()\n\nplt.show()\n\n# + [markdown] id=\"_co70oX3DUhs\" colab_type=\"text\"\n# 為數據集準備高解析率的資料產生器\n\n# + id=\"hrqjAXa-DUhs\" colab_type=\"code\" colab={}\n# This was previously set to 6 (one point per hour).\n# Now 3 (one point per 30 min).\nstep = 3\nlookback = 720 # Unchanged\ndelay = 144 # Unchanged\n\ntrain_gen = generator(float_data,\n lookback=lookback,\n delay=delay,\n min_index=0,\n max_index=200000,\n shuffle=True,\n step=step)\nval_gen = generator(float_data,\n lookback=lookback,\n delay=delay,\n min_index=200001,\n max_index=300000,\n step=step)\ntest_gen = generator(float_data,\n lookback=lookback,\n delay=delay,\n min_index=300001,\n max_index=None,\n step=step)\nval_steps = (300000 - 200001 - lookback) // 128\ntest_steps = (len(float_data) - 300001 - lookback) // 128\n\n# + [markdown] id=\"Jt1k09m9DUhv\" colab_type=\"text\"\n# This is our model, starting with two `Conv1D` layers and following-up with a `GRU` layer:\n\n# + id=\"WrE-mxV0HTjW\" colab_type=\"code\" colab={}\nmodel = Sequential()\nmodel.add(layers.Conv1D(32, 5, activation='relu',\n input_shape=(None, float_data.shape[-1])))\nmodel.add(layers.MaxPooling1D(3))\nmodel.add(layers.Conv1D(32, 5, activation='relu'))\nmodel.add(layers.GRU(32, dropout=0.1, recurrent_dropout=0.5))\nmodel.add(layers.Dense(1))\n\nmodel.summary()\n\nmodel.compile(optimizer=RMSprop(), loss='mae')\nhistory = model.fit_generator(train_gen,\n steps_per_epoch=500,\n epochs=20,\n validation_data=val_gen,\n validation_steps=val_steps)\n\n# + id=\"EHPAvagxHYlp\" colab_type=\"code\" colab={}\nloss = history.history['loss']\nval_loss = history.history['val_loss']\n\nepochs = range(len(loss))\n\nplt.figure()\n\nplt.plot(epochs, loss, 'bo', label='Training loss')\nplt.plot(epochs, val_loss, 'b', label='Validation loss')\nplt.title('Training and validation loss')\nplt.legend()\n\nplt.show()\n","repo_name":"jackliaoall/deep-learning-with-python-notebooks_exercises","sub_path":"6.4-sequence-processing-with-convnets.ipynb","file_name":"6.4-sequence-processing-with-convnets.ipynb","file_ext":"py","file_size_in_byte":11493,"program_lang":"python","lang":"en","doc_type":"code","stars":0,"dataset":"github-jupyter-script","pt":"40"}
+{"seq_id":"29830443818","text":"# # Multiple Linear Regression - Ejercicio - Publicidad y Ventas\n\n# **Contexto** \n# Este conjunto de datos contiene las ventas obtenidas, de acuerdo al monto invertido en publicidad en diferentes canales.\n#\n# **Contenido** \n# El conjunto de datos proviene de kaggle: [Advertising dataset](https://www.kaggle.com/datasets/tawfikelmetwally/advertising-dataset). \n# Contiene 200 renglones, con las siguientes columnas: \n#\n# | Variable | Definición | Valor |\n# | --------- | ---------------------------------------- | --------------- |\n# | TV | Inversión en anuncios por TV | miles de USD |\n# | Radio | Inversión en anuncios por radio | miles de USD |\n# | Newspaper | Inversión en anuncios por periódico | miles de USD |\n# | Sales | Ventas obtenidas **(variable objetivo)** | millones de USD | \n#\n# **Planteamiento del problema** \n# Se busca predecir que canales tienen mayor impacto en las ventas obtenidas.\n\n# +\n# Importar librerias\nimport pandas as pd\nimport numpy as np\nimport seaborn as sns\nimport matplotlib.pyplot as plt\n# %matplotlib inline\n\nfrom sklearn.preprocessing import StandardScaler\nfrom sklearn.model_selection import train_test_split\nfrom sklearn.linear_model import LinearRegression\nfrom sklearn import metrics\n# -\n\n# ## Cargar Datos\n\n# Importar los datos\ndf = pd.read_csv('Advertising.csv')\ndf.head()\n\n# Renombrar columnas\ndf.columns = ['renglón', 'tv', 'radio', 'periodico', 'ventas']\ndf.head()\n\n# ## Normalización\n\n# Variables independientes\nX = df[['tv', 'radio', 'periodico']]\nX.head()\n\n# Normalizar\nscaler = StandardScaler()\nX_adj = scaler.fit_transform(X)\nprint(X_adj)\n\n# Variable dependiente\ny = df['ventas']\ny.head()\n\nprint('X:', len(X_adj), 'y:', len(y))\n\n# ## Modelado\n\n# Conjunto de entrenamiento y pruebas\nX_train, X_test, y_train, y_test = train_test_split(X_adj, y, test_size=0.3, random_state=0)\n\nprint('X_train:', len(X_train), 'y_train:', len(y_train))\nprint('X_test:', len(X_test), 'y_test:', len(y_test))\n\n# Entrenamiento\nmodel = LinearRegression()\nmodel.fit(X_train,y_train)\n\n# Predicciones\nprediction = model.predict(X_test)\nprediction\n\n# Resultados\nprint(model.intercept_)\ncoef = pd.DataFrame(model.coef_, X.columns, columns=['coeficiente'])\ncoef\n\n# ## Evaluacion\n\nprint('MAE:', metrics.mean_absolute_error(y_test, prediction))\nprint('MSE:', metrics.mean_squared_error(y_test, prediction))\nprint('RMSE:', np.sqrt(metrics.mean_squared_error(y_test, prediction)))\nprint('R2:', metrics.r2_score(y_test, prediction))\n\nax1 = sns.kdeplot(y_test, color=\"r\")\nsns.kdeplot(prediction, color=\"b\", ax=ax1)\n\n\n","repo_name":"csameshiman/Axity_Curso_ML_2023","sub_path":"Ejercicios/Parte 02.Regresión/02-02-MLR-Ejemplo-Pub.ipynb","file_name":"02-02-MLR-Ejemplo-Pub.ipynb","file_ext":"py","file_size_in_byte":2640,"program_lang":"python","lang":"en","doc_type":"code","stars":1,"dataset":"github-jupyter-script","pt":"40"}
+{"seq_id":"14051035630","text":"import pandas as pd\nimport numpy as np\nimport matplotlib.pyplot as plt\nimport tensorflow as tf\nimport os\nfrom sklearn.model_selection import train_test_split\n\nfile_names = os.listdir('crop_part1')\n\n# +\nimport re\npattern = re.compile(r\"(\\d+)_(\\d+)_(\\d+)\")\n\ndf = pd.DataFrame(columns = ['Age', 'Gender', 'Ethnicity'])\ncount = 0\nfor i in file_names:\n data =list(pattern.search(i).groups())\n df.loc[count] = data\n count += 1\n \n \n#7881 and 7876 are bad values\n# AGE, Gender, Ethnicity\n# 0 is male, 1 is female\n# Race is an integer from 0 to 4, denoting White, Black, Asian, Indian, and Others (like Hispanic, Latino, Middle Eastern).\n# 0 = White\n# 1 = Black\n# 2 = Asian\n# 3 = Indian\n# 4 = Others\n\ndf['JPG'] = file_names\nimg_path = []\ncount = 0\nfor row in df.iterrows():\n path = \"crop_part1/\"+df.JPG.iloc[count]\n img_path.append(path)\n count += 1\ndf['img_path'] = img_path\ndf2 = df[['Age', 'Gender', 'Ethnicity', 'JPG', 'img_path']].copy()\ndf2.head()\n# -\n\ndf3 = pd.DataFrame(columns = ['Age_Range'])\nages = df2['Age'].tolist()\ncount = 0\nfor i in ages:\n i = int(i)\n if i == 0 or i == 1 or i == 2 or i == 3:\n df3.loc[count] = '0-2'\n elif i == 4 or i == 5 or i == 6 or i == 7:\n df3.loc[count] = '4-6'\n elif i == 8 or i == 9 or i == 10 or i == 11 or i == 12 or i == 13 or i == 14:\n df3.loc[count] = '8-13'\n elif i == 15 or i == 16 or i == 17 or i == 18 or i == 19 or i == 20 or i == 21 or i == 22 or i == 23 or i == 24:\n df3.loc[count] = '15-20'\n elif i == 25 or i == 26 or i == 27 or i == 28 or i == 29 or i == 30 or i == 31 or i == 32 or i == 33 or i == 34 or i == 35 or i == 36 or i == 37 :\n df3.loc[count] = '25-32'\n elif i == 38 or i == 39 or i == 40 or i == 41 or i == 42 or i == 43 or i == 44 or i == 45 or i == 46 or i == 47:\n df3.loc[count] = '38-43'\n elif i == 48 or i == 49 or i == 50 or i == 51 or i == 52 or i == 53 or i == 54 or i == 55 or i == 56 or i == 57 or i == 58 or i == 59:\n df3.loc[count] = '48-53'\n else: \n df3.loc[count] = '60+'\n count += 1\ndf3['Age_Range'].value_counts()\n\ndf4 = pd.concat([df2, df3], axis=1, join='inner')\nage_to_label_map = {\n '0-2' :0,\n '4-6' :1,\n '8-13' :2,\n '15-20':3,\n '25-32':4,\n '38-43':5,\n '48-53':6,\n '60+' :7\n}\ndf4['Age_Range_Num'] = df4['Age_Range'].apply(lambda Age_Range: age_to_label_map[Age_Range])\ndf4.head()\n\n# +\ngender_df = pd.concat([df4,pd.get_dummies(df4['Gender'])],axis = 1)\ngender_df = gender_df[['JPG','0','1']]\ngender_df.columns = ['JPG','Male','Female']\ngender_df.to_csv('gender.csv')\n\nage_df = pd.concat([df4,pd.get_dummies(df4['Age_Range'])],axis = 1)\nage_df = age_df[['JPG','0-2','8-13','15-20','25-32','38-43','4-6','48-53','60+']]\nage_df.to_csv('age.csv')\n\nethnicity_df = pd.concat([df4,pd.get_dummies(df4['Ethnicity'])],axis = 1)\nethnicity_df = ethnicity_df[['JPG','0','1','2','3','4']]\nethnicity_df.columns = ['JPG','White','Black','Asian','Indian','Other']\nethnicity_df.to_csv('ethnicity.csv')\n# -\n\n# # Rename Gender Files\n\n \n\n# +\n# organize dataset into a useful structure\nfrom os import makedirs\nfrom os import listdir\nfrom shutil import copyfile\nfrom random import seed\nfrom random import random\n\n# seed random number generator\nseed(1)\n# define ratio of pictures to use for validation\nsrc_directory = 'crop_part1/'\ndataset_home = 'dataset_male_vs_females2/'\nsubdirs = ['male/', 'female/']\nfor subdir in subdirs:\n newdir = dataset_home + subdir \n makedirs(newdir, exist_ok=True)\n \ncount = 0\nfor file in listdir(src_directory):\n src = src_directory + '/' + file\n if df2.iloc[count]['JPG'] == file and int(df2.iloc[count]['Gender']) == 0:\n dst = dataset_home + 'male/' + file\n copyfile(src, dst)\n elif df2.iloc[count]['JPG'] == file and int(df2.iloc[count]['Gender']) == 1:\n dst = dataset_home + 'female/' + file\n copyfile(src, dst)\n count += 1\n# -\n\ntrainM_names = os.listdir('dataset_male_vs_females2/male')\ntrainF_names = os.listdir('dataset_male_vs_females2/female')\nprint('Train Male: ',len(trainM_names))\nprint('Train Female: ',len(trainF_names))\n# Train Male: 4370\n# Train Female: 5406\n\ndf2.Gender.value_counts()\n\n# # Rename Age Files\n\n# +\n# organize dataset into a useful structure\nfrom os import makedirs\nfrom os import listdir\nfrom shutil import copyfile\nfrom random import seed\nfrom random import random\n\n# seed random number generator\nseed(1)\n# define ratio of pictures to use for validation\nsrc_directory = 'crop_part1/'\ndataset_home = 'dataset_age2/'\nsubdirs = ['0-2/', '4-6/', '8-13/', '15-20/', '25-32/', '38-43/', '48-53/', '60+/']\nfor subdir in subdirs:\n newdir = dataset_home + subdir \n makedirs(newdir, exist_ok=True)\n\ncount = 0\nfor file in listdir(src_directory):\n src = src_directory + '/' + file\n if df4.iloc[count]['JPG'] == file and int(df4.iloc[count]['Age_Range_Num']) == 0:\n dst = dataset_home + '0-2/' + file\n copyfile(src, dst)\n elif df4.iloc[count]['JPG'] == file and int(df4.iloc[count]['Age_Range_Num']) == 1:\n dst = dataset_home + '4-6/' + file\n copyfile(src, dst)\n elif df4.iloc[count]['JPG'] == file and int(df4.iloc[count]['Age_Range_Num']) == 2:\n dst = dataset_home + '8-13/' + file\n copyfile(src, dst)\n elif df4.iloc[count]['JPG'] == file and int(df4.iloc[count]['Age_Range_Num']) == 3:\n dst = dataset_home + '15-20/' + file\n copyfile(src, dst)\n elif df4.iloc[count]['JPG'] == file and int(df4.iloc[count]['Age_Range_Num']) == 4:\n dst = dataset_home + '25-32/' + file\n copyfile(src, dst)\n elif df4.iloc[count]['JPG'] == file and int(df4.iloc[count]['Age_Range_Num']) == 5:\n dst = dataset_home + '38-43/' + file\n copyfile(src, dst)\n elif df4.iloc[count]['JPG'] == file and int(df4.iloc[count]['Age_Range_Num']) == 6:\n dst = dataset_home + '48-53/' + file\n copyfile(src, dst)\n else:\n dst = dataset_home + '60+/' + file\n copyfile(src, dst)\n \n count+= 1\n\n# +\nnames0 = os.listdir('dataset_age2/0-2')\nnames1 = os.listdir('dataset_age2/4-6')\nnames2 = os.listdir('dataset_age2/8-13')\nnames3 = os.listdir('dataset_age2/15-20')\nnames4 = os.listdir('dataset_age2/25-32')\nnames5 = os.listdir('dataset_age2/38-43')\nnames6 = os.listdir('dataset_age2/48-53')\nnames7 = os.listdir('dataset_age2/60+')\n\n\nprint('0-2: ',len(names0))\nprint('4-6: ',len(names1))\nprint('8-13: ',len(names2))\nprint('15-20: ',len(names3))\nprint('25-32: ',len(names4))\nprint('38-43: ',len(names5))\nprint('48-53: ',len(names6))\nprint('60+: ',len(names7))\nprint('Total: ', str(len(names0)+len(names1)+len(names2)+len(names3)+len(names4)+len(names5)+len(names6)+len(names7)))\n# -\n\ndf4.Age_Range_Num.value_counts()\n\n# # Rename Ethnicity Files\n\ndf2.Ethnicity.value_counts()\n\n# +\n# organize dataset into a useful structure\nfrom os import makedirs\nfrom os import listdir\nfrom shutil import copyfile\nfrom random import seed\nfrom random import random\n\n# 0 = White\n# 1 = Black\n# 2 = Asian\n# 3 = Indian\n# 4 = Others\n\n# seed random number generator\nseed(1)\n# define ratio of pictures to use for validation\nsrc_directory = 'crop_part1/'\ndataset_home = 'dataset_ethnicity/'\nsubdirs = ['White/', 'Black/', 'Asian/', 'Indian/', 'Others/']\nfor subdir in subdirs:\n newdir = dataset_home + subdir \n makedirs(newdir, exist_ok=True)\n\ncount = 0\nfor file in listdir(src_directory):\n src = src_directory + '/' + file\n if df2.iloc[count]['JPG'] == file and int(df2.iloc[count]['Ethnicity']) == 0:\n dst = dataset_home + 'White/' + file\n copyfile(src, dst)\n elif df2.iloc[count]['JPG'] == file and int(df2.iloc[count]['Ethnicity']) == 1:\n dst = dataset_home + 'Black/' + file\n copyfile(src, dst)\n elif df2.iloc[count]['JPG'] == file and int(df2.iloc[count]['Ethnicity']) == 2:\n dst = dataset_home + 'Asian/' + file\n copyfile(src, dst)\n elif df2.iloc[count]['JPG'] == file and int(df2.iloc[count]['Ethnicity']) == 3:\n dst = dataset_home + 'Indian/' + file\n copyfile(src, dst)\n else:\n dst = dataset_home + 'Others/' + file\n copyfile(src, dst)\n count += 1\n\n# +\nW_names = os.listdir('dataset_Ethnicity/White')\nB_names = os.listdir('dataset_Ethnicity/Black')\nA_names = os.listdir('dataset_Ethnicity/Asian')\nI_names = os.listdir('dataset_Ethnicity/Indian')\nO_names = os.listdir('dataset_Ethnicity/Others')\n\n\nprint('White: ',len(W_names))\nprint('Black: ',len(B_names))\nprint('Asian: ',len(A_names))\nprint('Indian: ',len(I_names))\nprint('Others: ',len(O_names))\nprint('Total: ', str(len(W_names)+len(B_names)+len(A_names)+len(I_names)+len(O_names)))\n\n\n# +\n# White: 5264\n# Black: 405\n# Asian: 1553\n# Indian: 1451\n# Others: 1103\n# Total: 9776\n# -\n\n# # Split files into train, test and val\n# ## Gender\n\n# +\nimport os\nimport numpy as np\nimport shutil\nimport pandas as pd\n\ndef train_test_split(root_dir, classes_dir, processed_dir):\n print(\"########### Train Test Val Script started ###########\")\n #data_csv = pd.read_csv(\"DataSet_Final.csv\") ##Use if you have classes saved in any .csv file\n\n val_ratio = 0.20\n test_ratio = 0.20\n\n for cls in classes_dir:\n # Creating partitions of the data after shuffeling\n print(\"$$$$$$$ Class Name \" + cls + \" $$$$$$$\")\n src = processed_dir +\"/\" + cls # Folder to copy images from\n\n allFileNames = os.listdir(src)\n np.random.shuffle(allFileNames)\n train_FileNames, val_FileNames, test_FileNames = np.split(np.array(allFileNames),\n [int(len(allFileNames) * (1 - (val_ratio + test_ratio))),\n int(len(allFileNames) * (1 - val_ratio)),\n ])\n\n train_FileNames = [src + '//' + name for name in train_FileNames.tolist()]\n val_FileNames = [src + '//' + name for name in val_FileNames.tolist()]\n test_FileNames = [src + '//' + name for name in test_FileNames.tolist()]\n\n print('Total images: '+ str(len(allFileNames)))\n print('Training: '+ str(len(train_FileNames)))\n print('Validation: '+ str(len(val_FileNames)))\n print('Testing: '+ str(len(test_FileNames)))\n\n # # Creating Train / Val / Test folders (One time use)\n os.makedirs(root_dir + '/train//' + cls)\n os.makedirs(root_dir + '/val//' + cls )\n os.makedirs(root_dir + '/test//' + cls)\n # Copy-pasting images\n for name in train_FileNames:\n shutil.copy(name, root_dir + '/train//' + cls)\n\n for name in val_FileNames:\n shutil.copy(name, root_dir + '/val//' + cls)\n\n for name in test_FileNames:\n shutil.copy(name, root_dir + '/test//' + cls)\n\n print(\"########### Train Test Val Script Ended ###########\")\n\n\n# -\n\nnew_direct = 'Gender_Images2'\nclasses = ['male', 'female']\nprocessed_direct = 'dataset_male_vs_females2/'\ntrain_test_split(new_direct, classes, processed_direct)\n\n# ## Age\n\n\n\nnew_direct2 = 'Age_Images2'\nclasses2 = ['0-2/', '4-6/', '8-13/', '15-20/', '25-32/', '38-43/', '48-53/', '60+/']\nprocessed_direct2 = 'dataset_age2/'\ntrain_test_split(new_direct2, classes2, processed_direct2)\n\n# # Ethnicity\n\n# +\n# Race is an integer from 0 to 4, denoting White, Black, Asian, Indian, and Others (like Hispanic, Latino, Middle Eastern).\n# 0 = White\n# 1 = Black\n# 2 = Asian\n# 3 = Indian\n# 4 = Others\n\nnew_direct3 = 'Ethnicity_Images'\nclasses3 = ['White','Black','Asian','Indian','Others']\nprocessed_direct3 = 'dataset_ethnicity/'\n\ntrain_test_split(new_direct3, classes3, processed_direct3)\n# -\n\n\n\n\n","repo_name":"NarainMandyam/GenderEthnicityAge-FaceClassifier","sub_path":"fileFormatting.ipynb","file_name":"fileFormatting.ipynb","file_ext":"py","file_size_in_byte":11659,"program_lang":"python","lang":"en","doc_type":"code","stars":0,"dataset":"github-jupyter-script","pt":"40"}
+{"seq_id":"36070262117","text":"# # Создание столбцов на ходу\n# - Лена - разработчик\n# - Ваня - дизайнер\n# - Леша - редактор\n\nimport pandas as pd\n\ndf = pd.DataFrame({'dicts': [{'Лена': 1, 'Ваня': 2}, {'Ваня': 3, 'Леша': 5}, {'Настя': 3}]})\ndf\n\n\ndef departments(row):\n if 'Ваня' in row['dicts']:\n row['разработка'] = row['dicts']['Ваня']\n \n if 'Настя' in row['dicts']:\n row['финансы'] = row['dicts']['Настя']\n \n return row\n\n\ndf = df.apply(departments, axis=1)\ndf\n\nstaff = {\n 'Лена': 'разработчик',\n 'Ваня': 'дизайнер',\n 'Настя': 'редактор',\n 'Кеша': 'директорат',\n}\n\n\ndef new_columns(row):\n for name, value in row['dicts'].items():\n if name in staff:\n row[staff[name]] = value\n \n return row\n\n\ndf = df.apply(new_columns, axis=1)\ndf\n","repo_name":"vn322/Netology_DS_course","sub_path":"NumPy, pandas, MPL/3.6.pandas_functions_and_rows.ipynb","file_name":"3.6.pandas_functions_and_rows.ipynb","file_ext":"py","file_size_in_byte":937,"program_lang":"python","lang":"en","doc_type":"code","stars":1,"dataset":"github-jupyter-script","pt":"40"}
+{"seq_id":"8078350712","text":"# +\nimport cv2\nimport numpy as np\nimport pandas as pd\nfrom tqdm.notebook import tqdm\nimport matplotlib.pyplot as plt\n# %matplotlib inline\n\nimport os\nimport sys\nfrom PIL import Image \nfrom glob import glob\nfrom pathlib import Path\nimport imageio.v2 as imageio\nfrom datetime import datetime\nfrom typing import Union, List, Optional, Tuple\n\nsys.path.append(str(Path().absolute().parent))\nimport plotting\nimport importlib\nimportlib.reload(plotting)\n# from plotting import overlay_masks_on_data\n# -\n\n# # Resources\n# - [5 Ways to Make Histopathology Image Models More Robust to Domain Shifts](https://towardsdatascience.com/5-ways-to-make-histopathology-image-models-more-robust-to-domain-shifts-323d4d21d889)\n#\n# # Questions\n# - Is the transparency/alpha information (RGBA vs RGB) relevant to solve the problem?\n# - The images were captured with different scanners, inducing image variations?\n# - How patches were sampled? There's some overlapping between them?\n# - How to reconstruct the patches?\n# - How to overlap the reconstructed image (from patches) with the mask?\n\n# # Helpers\n\n# +\nbins = np.linspace(0, 255, 52)\n\n\ndef exists(obj):\n return obj is not None\n\n\ndef print_img_info(img: np.ndarray, desc: str = \"\"):\n print(\n desc,\n f\"min: {img.min()}\",\n f\"max: {img.max() }\",\n f\"mean: {img.mean():.2f}\",\n f\"ndim: {img.ndim}\",\n f\"channels: {img.shape[-1]}\",\n )\n\n\ndef read_image(image_path: Union[str, Path]) -> np.ndarray:\n \"\"\"Reads an image from a path and converts it to RGB format.\"\"\"\n if isinstance(image_path, Path):\n image_path = str(image_path)\n image = cv2.cvtColor(cv2.imread(image_path), cv2.COLOR_BGR2RGB)\n return image.astype(np.float32)\n\n\ndef read_mask(mask_path: Union[str, Path]) -> np.ndarray:\n \"\"\"Reads a mask from a path and transform it to binary.\"\"\"\n if isinstance(mask_path, Path):\n mask_path = str(mask_path)\n mask = cv2.imread(mask_path, 0) / 255.0\n assert_mask_is_binary(mask)\n return mask.astype(np.float32)\n\n\ndef assert_mask_is_binary(mask: np.ndarray):\n \"\"\"Counts all the pixels different to zero and one to check if binary.\"\"\"\n assert (\n np.count_nonzero((mask != 0) & (mask != 1)) == 0\n ), f\"Mask is not binary. Unique values: {np.unique(mask)}\"\n\n\ndef read_images_grid(\n patches_paths: Union[List[str], List[Path]]\n) -> List[List[np.ndarray]]:\n patches = []\n for i, patches_paths_row in enumerate(patches_paths):\n images_row = []\n for j, patch_path in enumerate(patches_paths_row):\n image = read_image(patch_path) if exists(patch_path) else np.zeros((1024, 1024, 3))\n images_row.append(image)\n patches.append(images_row)\n return patches\n\n\ndef crop_black_frames_from_image(img: np.ndarray) -> np.ndarray: \n positions = np.nonzero(img)\n top = positions[0].min()\n bottom = positions[0].max()\n left = positions[1].min()\n right = positions[1].max()\n return img[top:bottom, left:right]\n\n\ndef sort_paths_from_idxs(\n patches_paths: Union[List[str], List[Path]]\n) -> List[List[Union[str, Path]]]:\n \"\"\"Creates a grid of paths from a list of paths based on their ids.\"\"\"\n if isinstance(patches_paths[0], str):\n patches_paths = [Path(patch_path) for patch_path in patches_paths]\n idxs = [patch_path.stem.split(\"_\")[-2:] for patch_path in patches_paths]\n idxs = [[int(i[0]), int(i[1])] for i in idxs]\n\n # Create the grid.\n (y_min, x_min), (y_max, x_max) = np.min(idxs, axis=0), np.max(idxs, axis=0)\n sorted_paths, missing_patches_idxs = [], []\n for j in range(y_max + 1):\n row_paths = []\n for i in range(x_max + 1):\n if [j, i] not in idxs:\n missing_patches_idxs.append([j, i])\n row_paths.append(None)\n else:\n row_paths.append(patches_paths[idxs.index([j, i])])\n sorted_paths.append(row_paths)\n\n return sorted_paths, missing_patches_idxs\n\n\n\n# -\n\n# # Exploration\n#\n# Create variables pointing to the directories holding the histopathological data\n\n# +\ndset_dir = \"/data/histopathology/TCGA/\"\ndset_dir = Path(dset_dir)\n\npatches_dir = dset_dir / \"patches\"\nmasks_dir = dset_dir / \"masks\"\n\n# In the patches directory are more directories and in the masks directory are png files.\npatches_dirs = sorted(patches_dir.iterdir())\nmasks_paths = sorted(masks_dir.iterdir())\n# -\n\n# ## How to read histopathological images\n#\n# The first thing to check is how are these images codified. Read them with different libraries.\n\n# +\n# Convert to RGB for cv2 images.\nfirst_patch_path = str(sorted(patches_dirs[0].iterdir())[0])\ncv_img = cv2.cvtColor(cv2.imread(first_patch_path), cv2.COLOR_BGR2RGB)\ncv_img_unchanged = cv2.cvtColor(cv2.imread(first_patch_path, -1), cv2.COLOR_BGRA2RGBA)\nio_img = imageio.imread(first_patch_path)\npil_img = np.array(Image.open(first_patch_path))\n\nprint_img_info(cv_img, desc=\"Standard cv imread\")\nprint_img_info(cv_img_unchanged, desc=\"Unchanged cv imread\")\nprint_img_info(io_img, desc=\"Imageio imread\")\nprint_img_info(pil_img, desc=\"PIL imread\")\n\nassert np.array_equal(cv_img_unchanged, io_img) and np.array_equal(\n cv_img_unchanged, pil_img\n), \"Images are not equal.\"\n\n# -\n\n# Imageio, PIL and OpenCV (if specified) read the images without changing their intensity values. The images seem to be codified in RGB with transparency. *Is the transparency information relevant to solve the problem?*\n#\n# Plot the different images to check if there are noticeable differences.\n\nfig, ax = plt.subplots(1, 4, figsize=(40, 10))\nax[0].imshow(cv_img)\nax[0].set_title(\"Standard cv imread\")\nax[1].imshow(cv_img_unchanged)\nax[1].set_title(\"Unchanged cv imread\")\nax[2].imshow(io_img)\nax[2].set_title(\"Imageio imread\")\nax[3].imshow(pil_img)\nax[3].set_title(\"PIL imread\")\n\n\n# There are no noticeable visual differences. Check the mean intensity for the last channel in the images with transparency\n\nfig, ax = plt.subplots(1, 4, figsize=(20, 5))\nax[0].set_title(\"Red channel\")\nax[0].hist(cv_img_unchanged[..., 0].ravel(), density=True, bins=bins)\nax[1].set_title(\"Green channel\")\nax[1].hist(cv_img_unchanged[..., 1].ravel(), density=True, bins=bins)\nax[2].set_title(\"Blue channel\")\nax[2].hist(cv_img_unchanged[..., 2].ravel(), density=True, bins=bins)\nax[3].set_title(\"Alpha channel\")\nax[3].hist(cv_img_unchanged[..., 3].ravel(), density=True, bins=bins)\n\n\n# Might be a bit hasty, but judging by the intensity distribution in the transparency channel, you could use the images with their RGB coding and not RGBA. In fact, if we look at the minimum, maximum and average values we can corroborate that all the information is located in a very small interval.\n\nprint(\n \"Alpha channel ->\",\n \"min:\",\n cv_img_unchanged[..., 3].min(),\n \"max:\",\n cv_img_unchanged[..., 3].max(),\n \"mean:\",\n cv_img_unchanged[..., 3].mean(),\n)\n\n\n# Take more samples and see if the same conclusions can be drawn\n\n# +\npatches_paths = []\nfor patch_dir in patches_dirs:\n patches_paths.extend(sorted(patch_dir.iterdir()))\n\n\nnp.random.seed(4321)\nrandom_patches_paths = np.random.choice(patches_paths, size=200, replace=False)\n\npatches = []\nfor random_patch_path in random_patches_paths:\n patch = cv2.cvtColor(cv2.imread(str(random_patch_path), -1), cv2.COLOR_BGRA2RGBA)\n patches.append(patch)\n\n\npatches = np.array(patches)\nprint_img_info(patches, desc=\"Patches\")\n\n# -\n\n# Plot five patches to see the difference\n\n# plot 5 random patches.\nfig, ax = plt.subplots(1, 5, figsize=(20, 5))\nfor i in range(5):\n ax[i].imshow(patches[i])\n ax[i].set_title(f\"Patch {i}\")\n\n\n# Plot the intensity distribution for each channel of the `patches` array. Compare it against the normalized histogram of the previous image\n\nfig, ax = plt.subplots(1, 4, figsize=(20, 5))\nax[0].set_title(\"Red channel\")\nax[0].hist(patches[..., 0].ravel(), density=True, bins=bins)\nax[0].hist(\n cv_img_unchanged[..., 0].ravel(),\n color=\"red\",\n density=True,\n histtype=\"step\",\n bins=bins,\n)\nax[1].set_title(\"Green channel\")\nax[1].hist(patches[..., 1].ravel(), density=True, bins=bins)\nax[1].hist(\n cv_img_unchanged[..., 1].ravel(),\n color=\"red\",\n density=True,\n histtype=\"step\",\n bins=bins,\n)\nax[2].set_title(\"Blue channel\")\nax[2].hist(patches[..., 2].ravel(), density=True, bins=bins)\nax[2].hist(\n cv_img_unchanged[..., 2].ravel(),\n color=\"red\",\n density=True,\n histtype=\"step\",\n bins=bins,\n)\nax[3].set_title(\"Alpha channel\")\nax[3].hist(patches[..., 3].ravel(), density=True, bins=bins, label=\"200 random patches\")\nax[3].hist(\n cv_img_unchanged[..., 3].ravel(),\n color=\"red\",\n density=True,\n histtype=\"step\",\n bins=bins,\n label=\"1 patch\",\n)\nax[3].legend()\n\n\n# *Distributions look very similar for one or multiple randomly sampled patches. Therefore, it seems that the alpha channel holds a lot of repeated information*. \n\n# ## Defining the splits\n#\n# The dataset has some samples without masks that are meant to be used for test. The samples that have a patches directory and a mask are for training. Compare the ids of the patches directories and the masks to find which are the test samples.\n\n# +\nmask_sample_ids = [mask_path.stem[:-5] for mask_path in masks_paths]\npatches_sample_ids = [patch_dir.name for patch_dir in patches_dirs]\ntest_sample_ids = sorted(list(set(patches_sample_ids).difference(mask_sample_ids)))\ntrain_sample_ids = sorted(list(set(patches_sample_ids).intersection(mask_sample_ids)))\n\nprint(f\"Number of samples with masks: {len(train_sample_ids)} \\n\")\nprint(f\"Samples ids without masks:\", *test_sample_ids, sep=\"\\n- \")\n\n# -\n\n# ## Error checking at upload\n#\n# Iterate over the training and test samples directories to make sure they have patches uploaded and that no mask exist for the test samples\n#\n# ### Train split\n#\n# There are six samples without images inside the patches directories. Aside from that, everything seems fine.\n\nfor train_sample_id in train_sample_ids:\n\n sample_mask_path = masks_dir / f\"{train_sample_id}_mask.png\"\n\n sample_patches_dir = patches_dir / train_sample_id\n sample_patches_paths = sorted(sample_patches_dir.iterdir())\n sample_num_patches = len(sample_patches_paths)\n\n if sample_num_patches == 0:\n print(f\"{train_sample_id} do not have patches\")\n\n if not sample_mask_path.exists():\n print(f\"{train_sample_id} do not have a mask\")\n\n\n# ### Test split\n#\n# Everything seems ordered for the samples in the test split, i.e. all the patch directories contain image and there are no masks.\n\nfor test_sample_id in test_sample_ids:\n\n sample_mask_path = masks_dir / f\"{test_sample_id}_mask.png\"\n\n sample_patches_dir = patches_dir / test_sample_id\n sample_patches_paths = sorted(sample_patches_dir.iterdir())\n sample_num_patches = len(sample_patches_paths)\n\n if sample_num_patches == 0:\n print(f\"{test_sample_id} do not have patches\")\n\n if sample_mask_path.exists():\n print(f\"{test_sample_id} has a mask\")\n\n\n# \n\n# ## Image reconstruction: how to overlap the masks and patches\n#\n# ### Unusual things\n\n# #### Square grid\n#\n# At first, I believed that the patches should make a squared grid. Nevertheless, taking a look at the samples from the training split, this idea quickly goes away.\n\nfor train_sample_id in train_sample_ids:\n\n sample_patches_dir = patches_dir / train_sample_id\n sample_patches_paths = sorted(sample_patches_dir.iterdir())\n sample_num_patches = len(sample_patches_paths)\n\n if sample_num_patches > 0:\n if not np.sqrt(sample_num_patches).is_integer():\n print(\n f\"{train_sample_id}: Number of patches is not a perfect square (n={sample_num_patches}).\"\n )\n\n\n# #### Missing patches\n\n# +\ntoday_str = datetime.now().strftime(\"%Y%m%d\")\nresults_dir = Path(f\"../report/{today_str}/\")\nresults_dir.mkdir(exist_ok=True, parents=True)\n\nreport_dict = {\n \"sample_id\": [],\n \"mask_shape\": [],\n \"reconstructed_image_shape\": [],\n \"num_patches\": [],\n \"num_missing_patches\": [],\n \"missing_patches_idxs\": [],\n}\n\npbar = tqdm(train_sample_ids)\nfor sample_id in pbar:\n\n pbar.set_description(f\"Processing {sample_id}\")\n\n sample_patches_dir = patches_dir / sample_id\n sample_patches_paths = sorted(sample_patches_dir.iterdir())\n\n if len(sample_patches_paths) > 0:\n sorted_sample_patches_paths, missing_patches_idxs = sort_paths_from_idxs(\n sample_patches_paths\n )\n sorted_sample_patches = read_images_grid(sorted_sample_patches_paths)\n reconstructed_sample_image = np.vstack(\n [np.hstack(row) for row in sorted_sample_patches]\n ).astype(np.float32)\n\n sample_mask_path = masks_dir / f\"{sample_id}_mask.png\"\n sample_mask = read_mask(sample_mask_path)\n\n # Process the reconstructed image.\n reconstructed_sample_image = crop_black_frames_from_image(reconstructed_sample_image)\n reconstructed_sample_image = reconstructed_sample_image / 255.0\n \n # Resize mask to cropped reconstructed image.\n height, width = reconstructed_sample_image.shape[:2]\n resized_sample_mask = cv2.resize(sample_mask, (width, height))\n data_with_mask = plotting.overlay_masks_on_data(\n np.expand_dims(reconstructed_sample_image, axis=0),\n np.expand_dims(resized_sample_mask, axis=-1),\n true_mask_with_contours=False,\n ).clip(0, 1)\n\n # Save results.\n sample_results_dir = results_dir / sample_id\n sample_results_dir.mkdir(exist_ok=True)\n\n report_dict[\"sample_id\"].append(sample_id)\n report_dict[\"num_missing_patches\"].append(len(missing_patches_idxs))\n report_dict[\"missing_patches_idxs\"].append(\n \", \".join([str(i) for i in missing_patches_idxs])\n )\n report_dict[\"mask_shape\"].append(sample_mask.shape)\n report_dict[\"reconstructed_image_shape\"].append(\n reconstructed_sample_image.shape\n )\n report_dict[\"num_patches\"].append(len(sample_patches_paths))\n\n plt.figure(figsize=(20, 20))\n plt.imshow(data_with_mask[0])\n for j, i in missing_patches_idxs:\n y, x = 512 + (1024 * j), 350 + (1024 * i)\n plt.annotate(\n f\"$[{j}, {i}]$\",\n xy=(x, y),\n color=\"yellow\",\n fontsize=16,\n )\n plt.savefig(\n str(sample_results_dir / \"reconstructed_image_with_resize_mask.png\"),\n bbox_inches=\"tight\",\n )\n plt.close()\n\n else:\n\n sample_mask_path = masks_dir / f\"{sample_id}_mask.png\"\n sample_mask = read_mask(sample_mask_path)\n\n report_dict[\"sample_id\"].append(sample_id)\n report_dict[\"num_missing_patches\"].append(None)\n report_dict[\"missing_patches_idxs\"].append(None)\n report_dict[\"mask_shape\"].append(sample_mask.shape)\n report_dict[\"reconstructed_image_shape\"].append(None)\n report_dict[\"num_patches\"].append(0)\n\nreport_df = pd.DataFrame(report_dict)\nreport_df.to_excel(results_dir / \"report.xlsx\", index=False)\n\n# -\n\nreport_df\n\n\n# # Quitar TCGA-55-8207-01Z-00-DX1.2dafc442-f927-4b0d-b197-cc8c5f86d0fc\n\n\n","repo_name":"Sangohe/tcga-pathology-lung","sub_path":"notebooks/data-analysis.ipynb","file_name":"data-analysis.ipynb","file_ext":"py","file_size_in_byte":15251,"program_lang":"python","lang":"en","doc_type":"code","stars":0,"dataset":"github-jupyter-script","pt":"44"}
+{"seq_id":"10490091818","text":"# Import required python libraries and add some formatting options\n\n# +\nfrom math import sqrt\n\nimport numpy as np\nimport pandas as pd\nfrom scipy.optimize import minimize\nimport matplotlib.pyplot as plt\n\nfloat_formatter = lambda x: \"%.4f\" % x\nnp.set_printoptions(formatter={'float_kind':float_formatter})\n\n\n# -\n\n# Define a function that calculates optimal portfolio by minimising the volatility for a given level of risk\n\ndef optimise_portfolio(target_mu=0.045, corr_factor=1):\n \"\"\" Returns a string decribing optimial portfolio allocation\n \n target_mu: float, target portfolio return\n corr_factor: float, value to scale correlation up by, useful for q1.3\n \n \"\"\"\n \n mu = np.matrix([\n [0.02], \n [0.07], \n [0.15], \n [0.20],\n ])\n\n sigma = np.diag([\n 0.05, \n 0.12, \n 0.17, \n 0.25,\n ])\n\n correlation = np.matrix([\n [1.00, corr_factor * 0.30, corr_factor * 0.30, corr_factor * 0.30],\n [corr_factor * 0.30, 1.00, corr_factor * 0.60, corr_factor * 0.60],\n [corr_factor * 0.30, corr_factor * 0.60, 1.00, corr_factor * 0.60],\n [corr_factor * 0.30, corr_factor * 0.60, corr_factor * 0.60, 1.00],\n ])\n\n covariance = sigma @ correlation @ sigma\n\n start_weights = np.matrix([\n [0.25], \n [0.25], \n [0.25], \n [0.25],\n ])\n\n def calc_sigma(my_weights):\n return sqrt(my_weights.T @ covariance @ my_weights)\n\n r = minimize(calc_sigma, start_weights, tol=0.000000001,\n constraints=(\n {\n 'type': 'eq', \n 'fun': lambda weights: 1.0 - np.sum(weights)\n },\n {\n 'type': 'eq', \n 'fun': lambda weights: target_mu - (weights.T @ mu).item((0, 0))\n }\n ))\n \n optimal_weights = r.x\n pfo_mu = (optimal_weights.T @ mu).item((0, 0))\n pfo_risk = sqrt((optimal_weights.T @ covariance @ optimal_weights).item((0, 0)))\n \n return_value = 'Target:{}'.format(target_mu)\n return_value += ', Optimal Weights:{}'.format(optimal_weights)\n return_value += ', Mu:{:.4f}'.format(pfo_mu)\n return_value += ', SD:{:.4f}'.format(pfo_risk)\n return print(return_value)\n\n\noptimise_portfolio(0.045)\noptimise_portfolio(0.045, corr_factor=1)\noptimise_portfolio(0.045, corr_factor=1.25)\noptimise_portfolio(0.045, corr_factor=1.5)\n\n\ndef optimise_portfolio(risk_free_rate=0.005, corr_factor=1):\n\n mu = np.matrix([\n [0.02], \n [0.07], \n [0.15], \n [0.20],\n ])\n\n sigma = np.diag([\n 0.05, \n 0.12, \n 0.17, \n 0.25,\n ])\n\n correlation = np.matrix([\n [1.00, corr_factor * 0.30, corr_factor * 0.30, corr_factor * 0.30],\n [corr_factor * 0.30, 1.00, corr_factor * 0.60, corr_factor * 0.60],\n [corr_factor * 0.30, corr_factor * 0.60, 1.00, corr_factor * 0.60],\n [corr_factor * 0.30, corr_factor * 0.60, corr_factor * 0.60, 1.00],\n ])\n\n covariance = sigma @ correlation @ sigma\n\n start_weights = np.matrix([\n [0.25], \n [0.25], \n [0.25], \n [0.25],\n ])\n\n def calc_sharpe(my_weights):\n pf_ret = (my_weights.T @ mu).item(0, 0)\n excess_return = pf_ret - risk_free_rate\n pf_sd = sqrt(my_weights.T @ covariance @ my_weights)\n sharpe = excess_return / pf_sd\n return -sharpe\n\n r = minimize(calc_sharpe, start_weights, tol=0.000000001,\n constraints=(\n {\n 'type': 'eq', \n 'fun': lambda weights: 1.0 - np.sum(weights)\n },\n ))\n \n print('')\n print(r.message)\n optimal_weights = r.x\n pfo_mu = (optimal_weights.T @ mu).item((0, 0))\n pfo_excess_return = pfo_mu - risk_free_rate\n pfo_risk = sqrt((optimal_weights.T @ covariance @ optimal_weights).item((0, 0)))\n pfo_sharpe = pfo_excess_return / pfo_risk\n \n print('{:.4f} \\t& {:.4f} \\t& {:.4f} \\t& {:.4f} \\t& {:.4f} \\t& {:.4f} \\t& {:.4f} \\t& {:.4f} \\\\\\\\'.format(risk_free_rate, optimal_weights[0], optimal_weights[1], \n optimal_weights[2], optimal_weights[3], pfo_mu, pfo_risk, pfo_sharpe))\n \n return_value = ' Risk Free:{}'.format(risk_free_rate)\n return_value += '\\n Optimal Weights:{}'.format(optimal_weights)\n return_value += '\\n Mu:{:.4f}'.format(pfo_mu)\n return_value += '\\n SD:{:.4f}'.format(pfo_risk)\n return_value += '\\n Sharpe: {:.4f}'.format(pfo_sharpe)\n return print(return_value)\n\n\noptimise_portfolio(0)\noptimise_portfolio(0.0050)\noptimise_portfolio(0.0075)\noptimise_portfolio(0.0100)\noptimise_portfolio(0.0150)\noptimise_portfolio(0.0175)\n\n\n# Inverse optimisation problem where we generate weights randomly\n\ndef question_2_part_2():\n n_weights = 1500\n \n mu = np.matrix([\n [0.02], \n [0.07], \n [0.15], \n [0.20],\n ])\n \n sigma = np.diag([\n 0.05, \n 0.12, \n 0.17, \n 0.25,\n ])\n \n # draw (time_steps x n_steps) random variables\n r = np.random.randn(4, n_weights)\n print('Sample weights: \\n{}'.format(r[:,:5]))\n \n # find the sum of the four random weights\n s = np.sum(r, axis=0)\n print('Sample sum weights: \\n{}'.format(s[:5]))\n \n # normalise each weight by dividing by sum of weights to ensure sum of weights add up to 1\n random_weights = (r.T / s[:, None]).T # slice with None and combine with broadcasting\n print('Sample normalised weights: \\n{}'.format(random_weights[:,:5]))\n \n # check random weights add up to 1\n # total number of random portfolios = 1500\n # ensure that range of weight values look reasonable\n check_random_weights = pd.DataFrame(random_weights.T)\n check_random_weights['sum'] = check_random_weights.sum(axis=1)\n print('Checking random weight statistics:\\n{}'.format(check_random_weights.describe()))\n \n rand_pf_mu = random_weights.T @ mu\n print('Sample random portfolio expected returns:\\n{}'.format(rand_pf_mu[:5]))\n \n correlation = np.matrix([\n [1.00, 0.30, 0.30, 0.30],\n [0.30, 1.00, 0.60, 0.60],\n [0.30, 0.60, 1.00, 0.60],\n [0.30, 0.60, 0.60, 1.00],\n ])\n\n covariance = sigma @ correlation @ sigma\n \n a = random_weights.T @ covariance @ random_weights\n rand_pf_sigma = np.sqrt(a.diagonal().T)\n print('Sample random portfolio volatility:\\n{}'.format(rand_pf_sigma[:5]))\n \n # combine results into a dataframe so that we can more easily visualise\n df = pd.concat([\n pd.DataFrame(rand_pf_mu, columns=['pf_mu']), \n pd.DataFrame(rand_pf_sigma, columns=['pf_sigma'])\n ], axis=1)\n # display the first five elements\n print(df.head())\n \n return df.plot(x=['pf_sigma'], y=['pf_mu'], kind='scatter', \n xticks=np.arange(0, 0.25, .05), yticks=np.arange(0, 0.25, .05),\n alpha=0.2, figsize=(12, 8), title='Efficient Frontier generated by inverse optimisation'\n )\n\n\nquestion_2_part_2()\n\n# +\ncorr_bump = -0.5\n\ndef plot_for_corr_bump(corr):\n \n n_weights = 1500\n \n mu = np.matrix([\n [0.02], \n [0.07], \n [0.15], \n [0.20],\n ])\n sigma = np.diag([\n 0.05, \n 0.12, \n 0.17, \n 0.25,\n ])\n correlation = np.matrix([\n [1.00, corr, corr, corr],\n [corr, 1.00, corr, corr],\n [corr, corr, 1.00, corr],\n [corr, corr, corr, 1.00],\n ])\n \n # draw (time_steps x n_steps) random variables\n r = np.random.randn(4, n_weights)\n print('Sample weights: \\n{}'.format(r[:,:5]))\n \n # find the sum of the four random weights\n s = np.sum(r, axis=0)\n print('Sample sum weights: \\n{}'.format(s[:5]))\n \n # normalise each weight by dividing by sum of weights to ensure sum of weights add up to 1\n random_weights = (r.T / s[:, None]).T # slice with None and combine with broadcasting\n print('Sample normalised weights: \\n{}'.format(random_weights[:,:5]))\n\n covariance = sigma @ correlation @ sigma\n rand_pf_mu = random_weights.T @ mu\n temp_df = random_weights.T @ covariance @ random_weights\n rand_pf_sigma = np.sqrt(temp_df.diagonal().T)\n df = pd.concat([pd.DataFrame(rand_pf_mu, columns=['pf_mu']), pd.DataFrame(rand_pf_sigma, columns=['pf_sigma'])], axis=1)\n return df.plot(x=['pf_sigma'], y=['pf_mu'], kind='scatter', \n xticks=np.arange(0, 0.25, .05), yticks=np.arange(0, 0.25, .05),\n alpha=0.35, figsize=(12, 8), title='Average correlation = {}'.format(corr)\n )\n\n\n# -\n\nfor x in [-0.5, -0.25, 0, 0.25, 0.5, 0.75, 1]:\n plot_for_corr_bump(x);\n\n\n","repo_name":"yijxiang/quant-finance","sub_path":".ipynb_checkpoints/2019-08-29-optimal-portfolio-allocation-checkpoint.ipynb","file_name":"2019-08-29-optimal-portfolio-allocation-checkpoint.ipynb","file_ext":"py","file_size_in_byte":8653,"program_lang":"python","lang":"en","doc_type":"code","stars":0,"dataset":"github-jupyter-script","pt":"44"}
+{"seq_id":"9480801041","text":"# +\n#importing libraries\n\nimport pandas as pd\nimport numpy as np\n\nimport matplotlib.pyplot as plt\nimport seaborn as sns\n\nimport missingno as mno\n\nimport datetime\n\nplt.style.use('fivethirtyeight')\n# -\n\n#reading the data and also the computation time\n# %time data = pd.read_csv('data-1.csv')\n\n#shape of the dataset\ndata.shape\n\n#checking the columns present in the dataframe\ndata.columns\n\n# ## Data Cleaning\n\n# +\n#checking missing values present in the data\n#with missingno checking missing values for first 40 columns in a matrix form\n\nmno.bar(data.iloc[:, :40],\n color = 'orange',\n sort = 'ascending')\nplt.title('Checking missing values for first half of the data', fontsize= 17)\nplt.show()\n\n# +\n#checking missing values with missingno for second half\n\nmno.bar(data.iloc[:, 40:],\n sort= 'ascending')\nplt.title('Checking missing values for second half of data', fontsize= 17)\nplt.show()\n# -\n\n# #### Missing Values Imputation\n\n# +\n#imputing missing values for continuous varaibles\n\ndata['ShortPassing'].fillna(data['ShortPassing'].mean(), inplace = True)\ndata['Volleys'].fillna(data['Volleys'].mean(), inplace = True)\ndata['Dribbling'].fillna(data['Dribbling'].mean(), inplace = True)\ndata['Curve'].fillna(data['Curve'].mean(), inplace = True)\ndata['FKAccuracy'].fillna(data['FKAccuracy'], inplace = True)\ndata['LongPassing'].fillna(data['LongPassing'].mean(), inplace = True)\ndata['BallControl'].fillna(data['BallControl'].mean(), inplace = True)\ndata['HeadingAccuracy'].fillna(data['HeadingAccuracy'].mean(), inplace = True)\ndata['Finishing'].fillna(data['Finishing'].mean(), inplace = True)\ndata['Crossing'].fillna(data['Crossing'].mean(), inplace = True)\n\ndata['Weight'].fillna('200lbs', inplace = True)\ndata['Contract Valid Until'].fillna(2019, inplace = True)\ndata['Height'].fillna(\"5'11\", inplace = True)\ndata['Loaned From'].fillna('None', inplace = True)\ndata['Joined'].fillna('Jul 1, 2018', inplace = True)\ndata['Jersey Number'].fillna(8, inplace = True)\ndata['Body Type'].fillna('Normal', inplace = True)\ndata['Position'].fillna('ST', inplace = True)\ndata['Club'].fillna('No Club', inplace = True)\ndata['Work Rate'].fillna('Medium/ Medium', inplace = True)\ndata['Skill Moves'].fillna(data['Skill Moves'].median(), inplace = True)\ndata['Weak Foot'].fillna(3, inplace = True)\ndata['Preferred Foot'].fillna('Right', inplace = True)\ndata['International Reputation'].fillna(1, inplace = True)\ndata['Wage'].fillna('€200K', inplace = True)\n# -\n\npd.set_option('max_rows', 100)\ndata.isnull().sum()\n\n# +\n#imputing 0 for rest of missing value\n\ndata.fillna(0, inplace= True)\n# -\n\n#check missing values if present in the dataset\ndata.isnull().sum().sum()\n\n\n# ## Feature Engineering\n\n# +\n#creating new features by aggregrating similar features\n\ndef defending(data):\n return int(round((data[['Marking', 'StandingTackle', \n 'SlidingTackle']].mean()).mean()))\n\ndef general(data):\n return int(round((data[['HeadingAccuracy', 'Dribbling', 'Curve', \n 'BallControl']].mean()).mean()))\n\ndef mental(data):\n return int(round((data[['Aggression', 'Interceptions', 'Positioning', \n 'Vision','Composure']].mean()).mean()))\n\ndef passing(data):\n return int(round((data[['Crossing', 'ShortPassing', \n 'LongPassing']].mean()).mean()))\n\ndef mobility(data):\n return int(round((data[['Acceleration', 'SprintSpeed', \n 'Agility','Reactions']].mean()).mean()))\ndef power(data):\n return int(round((data[['Balance', 'Jumping', 'Stamina', \n 'Strength']].mean()).mean()))\n\ndef rating(data):\n return int(round((data[['Potential', 'Overall']].mean()).mean()))\n\ndef shooting(data):\n return int(round((data[['Finishing', 'Volleys', 'FKAccuracy', \n 'ShotPower','LongShots', 'Penalties']].mean()).mean()))\n\n\n# +\n#adding the above new features into the data\n\ndata['Defending'] = data.apply(defending, axis = 1)\ndata['General'] = data.apply(general, axis = 1)\ndata['Mental'] = data.apply(mental, axis = 1)\ndata['Passing'] = data.apply(passing, axis = 1)\ndata['Mobility'] = data.apply(mobility, axis = 1)\ndata['Power'] = data.apply(power, axis = 1)\ndata['Rating'] = data.apply(rating, axis = 1)\ndata['Shooting'] = data.apply(shooting, axis = 1)\n\n#check the column names in the data after adding new features\ndata.columns\n# -\n\ndata.head()\n\n# ## Data Visualization\n\n# +\n#visualize the new features(skills) created\n\nplt.rcParams['figure.figsize'] = (18,12)\nplt.subplot(2, 4, 1)\nsns.distplot(data['Defending'], color= 'red')\nplt.grid()\n\nplt.subplot(2, 4, 2)\nsns.distplot(data['General'], color= 'black')\nplt.grid()\n\nplt.subplot(2, 4, 3)\nsns.distplot(data['Mental'], color= 'red')\nplt.grid()\n\nplt.subplot(2, 4, 4)\nsns.distplot(data['Passing'], color= 'black')\nplt.grid()\n\nplt.subplot(2, 4, 5)\nsns.distplot(data['Mobility'], color= 'red')\nplt.grid()\n\nplt.subplot(2, 4, 6)\nsns.distplot(data['Power'], color= 'black')\nplt.grid()\n\nplt.subplot(2, 4, 7)\nsns.distplot(data['Rating'], color= 'red')\nplt.grid()\n\nplt.subplot(2, 4, 8)\nsns.distplot(data['Shooting'], color= 'black')\nplt.grid()\n\nplt.suptitle('Score distribution for different abilities')\nplt.show()\n\n# +\n#comparison of preferred foot for players\n\nplt.rcParams['figure.figsize'] = (8,3)\nsns.countplot(data['Preferred Foot'], palette = 'pink')\nplt.title('Most preferred foot for players', fontsize= 12)\nplt.show()\n\n# +\n# plotting a pie chart to represent share of international repuatation\n\nlabels = ['1', '2', '3', '4', '5'] #data['International Reputation'].index\nsizes = data['International Reputation'].value_counts()\ncolors = plt.cm.copper(np.linspace(0, 1, 5))\nexplode = [0.1, 0.1, 0.2, 0.5, 0.9]\n\nplt.rcParams['figure.figsize'] = (9, 9)\nplt.pie(sizes, labels = labels, colors = colors, explode = explode, shadow = True,)\nplt.title('International Repuatation for the Football Players', fontsize = 20)\nplt.legend()\nplt.show()\n# -\n\n# #### Lets check players with Internation rating of 5\n\ndata[data['International Reputation'] == 5][['Name','Nationality',\n 'Overall']].sort_values(by = 'Overall',\n ascending = False).style.background_gradient(cmap = 'magma')\n\n# +\n# plotting a pie chart to represent the share of week foot players\n\nlabels = ['5', '4', '3', '2', '1'] \nsize = data['Weak Foot'].value_counts()\ncolors = plt.cm.Wistia(np.linspace(0, 1, 5))\nexplode = [0, 0, 0, 0, 0.1]\n\nplt.pie(size, labels = labels, colors = colors, explode = explode, shadow = True, startangle = 90)\nplt.title('Distribution of Week Foot among Players', fontsize = 25)\nplt.legend()\nplt.show()\n\n# +\n# different positions acquired by the players \n\nplt.figure(figsize = (13, 15))\nax = sns.countplot(y = 'Position', data = data, palette = 'bone')\nax.set_ylabel(ylabel = 'Different Positions in Football', fontsize = 16)\nax.set_xlabel(xlabel = 'Count of Players', fontsize = 16)\nax.set_title(label = 'Comparison of Positions and Players', fontsize = 20)\nplt.show()\n\n\n# +\n# defining a function for cleaning the Weight data\n\ndef extract_value_from(value):\n out = value.replace('lbs', '')\n return float(out)\n\n# applying the function to weight column\n#data['value'] = data['value'].apply(lambda x: extract_value_from(x))\ndata['Weight'] = data['Weight'].apply(lambda x : extract_value_from(x))\n\n# plotting the distribution of weight of the players\nsns.distplot(data['Weight'], color = 'black')\nplt.title(\"Distribution of Players Weight\", fontsize = 15)\nplt.show()\n\n\n# +\n# defining a function for cleaning the wage column\n\ndef extract_value_from(column):\n out = column.replace('€', '')\n if 'M' in out:\n out = float(out.replace('M', ''))*1000000\n elif 'K' in column:\n out = float(out.replace('K', ''))*1000\n return float(out)\n\n\n# +\n# applying the function to the wage and value column\ndata['Value'] = data['Value'].apply(lambda x: extract_value_from(x))\ndata['Wage'] = data['Wage'].apply(lambda x: extract_value_from(x))\n\n# visualizing the data\nplt.rcParams['figure.figsize'] = (16, 5)\nplt.subplot(1, 2, 1)\nsns.distplot(data['Value'], color = 'violet')\nplt.title('Distribution of Value of the Players', fontsize = 15)\n\nplt.subplot(1, 2, 2)\nsns.distplot(data['Wage'], color = 'purple')\nplt.title('Distribution of Wages of the Players', fontsize = 15)\nplt.show()\n\n# +\n# Skill Moves of Players\n\nplt.figure(figsize = (10, 6))\nax = sns.countplot(x = 'Skill Moves', data = data, palette = 'pastel')\nax.set_title(label = 'Count of players on Basis of their skill moves', fontsize = 20)\nax.set_xlabel(xlabel = 'Number of Skill Moves', fontsize = 16)\nax.set_ylabel(ylabel = 'Count', fontsize = 16)\nplt.show()\n\n# +\n# To show Different Work rate of the players participating in the FIFA 2019\n\nplt.figure(figsize = (15, 5))\nplt.style.use('fivethirtyeight')\n\nsns.countplot(x = 'Work Rate', data = data, palette = 'hls')\nplt.title('Different work rates of the Players Participating in the FIFA 2019', fontsize = 20)\nplt.xlabel('Work rates associated with the players', fontsize = 16)\nplt.ylabel('count of Players', fontsize = 16)\nplt.xticks(rotation = 90)\nplt.show()\n\n# +\n#to show potential and overall scored of players\n\n\nplt.figure(figsize=(16, 4))\nplt.style.use('seaborn-paper')\n\nplt.subplot(1, 2, 1)\nx = data.Potential\nax = sns.distplot(x, bins = 58, kde = False, color = 'y')\nax.set_xlabel(xlabel = \"Player's Potential Scores\", fontsize = 10)\nax.set_ylabel(ylabel = 'Number of players', fontsize = 10)\nax.set_title(label = 'Histogram of players Potential Scores', fontsize = 15)\n\nplt.subplot(1, 2, 2)\ny = data.Overall\nax = sns.distplot(y, bins = 58, kde = False, color = 'y')\nax.set_xlabel(xlabel = \"Player's Overall Scores\", fontsize = 10)\nax.set_ylabel(ylabel = 'Number of players', fontsize = 10)\nax.set_title(label = 'Histogram of players Overall Scores', fontsize = 15)\nplt.show()\n\n# +\n#Overall scores of players with age distribution along with preferred foot\n\nplt.rcParams['figure.figsize'] = (20, 7)\nplt.style.use('seaborn-dark-palette')\n\nsns.boxplot(data['Overall'], data['Age'], hue = data['Preferred Foot'], palette = 'Greys')\nplt.title('Comparison of Overall Scores and age wrt Preferred foot', fontsize = 20)\nplt.show()\n\n# +\n#selecting the countries with highest number of players to compare their overall scores\n\ndata['Nationality'].value_counts().head(10).plot(kind = 'pie', cmap = 'inferno',\n startangle = 90, explode = [0, 0, 0, 0, 0, 0, 0, 0, 0.1, 0])\nplt.title('Countries having Highest Number of players', fontsize = 15)\nplt.axis('off')\nplt.show()\n\n# +\n#distribution of player weight's among the countries\n\nsome_countries = ('England', 'Germany', 'Spain', 'Argentina', 'France', 'Brazil', 'Italy', 'Columbia')\ndata_countries = data.loc[data['Nationality'].isin(some_countries) & data['Weight']]\n\ndata_countries\n# -\n\ndata_countries.columns\n\nplt.rcParams['figure.figsize'] = (15, 7)\nax = sns.violinplot(x = data_countries['Nationality'], y = data_countries['Weight'], palette = 'Reds')\nax.set_xlabel(xlabel = 'Countries', fontsize = 9)\nax.set_ylabel(ylabel = 'Weight in lbs', fontsize = 9)\nax.set_title(label = 'Distribution of Weight of players from different countries', fontsize = 20)\nplt.show()\n\n# +\n#Overall scores for Nations players\n\n\nsome_countries = ('England', 'Germany', 'Spain', 'Argentina', 'France', 'Brazil', 'Italy', 'Columbia')\ndata_countries = data.loc[data['Nationality'].isin(some_countries) & data['Overall']]\n\ndata_countries\n\n# +\n\nplt.rcParams['figure.figsize'] = (15, 7)\nax = sns.barplot(x = data_countries['Nationality'], y = data_countries['Overall'], palette = 'spring')\nax.set_xlabel(xlabel = 'Countries', fontsize = 9)\nax.set_ylabel(ylabel = 'Overall Scores', fontsize = 9)\nax.set_title(label = 'Distribution of overall scores of players from different countries', fontsize = 20)\nplt.show()\n\n# +\n#Nation's players and wages\n\nsome_countries = ('England', 'Germany', 'Spain', 'Argentina', 'France', 'Brazil', 'Italy', 'Columbia')\ndata_countries = data.loc[data['Nationality'].isin(some_countries) & data['Wage']]\n\nplt.rcParams['figure.figsize'] = (15, 7)\nax = sns.barplot(x = data_countries['Nationality'], y = data_countries['Wage'], palette = 'Purples')\nax.set_xlabel(xlabel = 'Countries', fontsize = 9)\nax.set_ylabel(ylabel = 'Wage', fontsize = 9)\nax.set_title(label = 'Distribution of Wages of players from different countries', fontsize = 15)\nplt.grid()\nplt.show()\n\n# +\n#Every player's reputation representing the Countries\n\n#Every Nations' Player and their International Reputation\n\nsome_countries = ('England', 'Germany', 'Spain', 'Argentina', 'France', 'Brazil', 'Italy', 'Columbia')\ndata_countries = data.loc[data['Nationality'].isin(some_countries) & data['International Reputation']]\n\nplt.rcParams['figure.figsize'] = (15, 7)\nax = sns.boxenplot(x = data_countries['Nationality'], y = data_countries['International Reputation'], palette = 'autumn')\nax.set_xlabel(xlabel = 'Countries', fontsize = 9)\nax.set_ylabel(ylabel = 'Distribution of reputation', fontsize = 9)\nax.set_title(label = 'Distribution of International Repuatation of players from different countries', fontsize = 15)\nplt.grid()\nplt.show()\n\n\n# +\n#Players reports function\n\ndef report(player):\n return data[data['Name'] == player][['Nationality','Club',\n 'Overall','Potential',\n 'Contract Valid Until','Wage',\n 'International Reputation']].reset_index(drop = True).T\n\n\n# -\n\nreport('L. Messi')\n\nreport('Cristiano Ronaldo')\n\nreport('Neymar Jr')\n","repo_name":"brianlobo394/FIFA18Players_Analysis","sub_path":"FIFA18_EDA.ipynb","file_name":"FIFA18_EDA.ipynb","file_ext":"py","file_size_in_byte":13639,"program_lang":"python","lang":"en","doc_type":"code","stars":0,"dataset":"github-jupyter-script","pt":"44"}
+{"seq_id":"21941201151","text":"import pandas as pd\nimport numpy as np\nimport seaborn as sns\nimport matplotlib.pyplot as plt\ndf=pd.read_csv(\"/home/jayesh/Desktop/cardio_train.csv\") \ndf.head()\nnp.isnan(df).sum()\nprint(type(df))\ndf.columns \n\nduplicates = len(df) - len(df.drop(['id'],axis=1).drop_duplicates())\ndf.drop(['id'],axis=1,inplace=True)\ndf.drop_duplicates(inplace=True)\nprint(f'{duplicates} duplicate records dropped.')\n\ndf.describe().T.round(2)\n\ntmp = pd.crosstab(df['cholesterol'],df['cardio'],normalize='index')\ntmp.reset_index()\ntmp.columns = ['not diseased','diseased']\nfig, ax = plt.subplots(1,1)\nsns.countplot(df['cholesterol'],order=list(tmp.index), ax=ax)\nplot2 = ax.twinx()\nsns.pointplot(tmp.index,tmp['diseased'],order=list(tmp.index),ax=plot2)\nfor patch in ax.patches:\n height = patch.get_height()\n ax.text(patch.get_x()+patch.get_width()/2,height,'{:.2f}{}'.format(height/len(df['cholesterol'])*100,'%'))\nplt.show()\n\nmp = pd.crosstab(df['gluc'],df['cardio'],normalize='index')\ntmp.reset_index()\ntmp.columns = ['not diseased','diseased']\nfig, ax = plt.subplots(1,1)\nsns.countplot(df['gluc'],order=list(tmp.index), ax=ax)\nplot2 = ax.twinx()\nsns.pointplot(tmp.index,tmp['diseased'],order=list(tmp.index),ax=plot2)\nfor patch in ax.patches:\n height = patch.get_height()\n ax.text(patch.get_x()+patch.get_width()/2,height,'{:.2f}{}'.format(height/len(df['gluc'])*100,'%'))\nplt.show()\n\n# +\nfig, (ax1,ax2) = plt.subplots(1,2, figsize=(20,10))\nsns.distplot(df['age'][df['cardio']==0], ax = ax1, color='green')\nsns.distplot(df['age'][df['cardio']==1], ax = ax1,color='coral')\nax1.set_title('Age Distribution')\nax1.legend()\n\nsns.distplot(df['age'][(df['gender']==1) & (df['cardio']==1)],ax = ax2,color='pink')\nsns.distplot(df['age'][(df['gender']==2) & (df['cardio']==1)],ax = ax2,color='blue')\nax2.set_title('Disease count distribution by gender, aged below 54.')\nplt.show()\n\n# +\nfig, (ax1) = plt.subplots(1,1, figsize=(10,10))\nsns.boxenplot(df['cardio'],(df['height']*0.0328084),ax=ax1)\nax1.set_title('Height / Diseased')\nplt.show()\n\n\n# -\n\nfrom pandas_profiling import ProfileReport\nprof = ProfileReport(df)\nprof.to_file(output_file='output.html')\n\ndf.corr()\n##\nX =df.drop('cardio', axis = 1) \ny=df['cardio']\n\n# +\nfrom sklearn.model_selection import train_test_split\n\nX_train, X_test, y_train, y_test = train_test_split(X, y,\n train_size=0.7, \n random_state=42,\n stratify=y)\n# -\n\nfrom sklearn.tree import DecisionTreeClassifier\nclf = DecisionTreeClassifier(criterion = 'entropy', min_samples_split=50)\n\nclf.fit(X_train, y_train)\n\ny_pred = clf.predict(X_test)\n\nfrom sklearn.metrics import accuracy_score\nprint('Accuracy Score on train data: ', accuracy_score(y_true=y_train, y_pred=clf.predict(X_train)))\nprint('Accuracy Score on test data: ', accuracy_score(y_true=y_test, y_pred=y_pred))\n\n#after feature engineering\n \n\n# decision tree\nfrom sklearn.tree import DecisionTreeClassifier\nclf = DecisionTreeClassifier(criterion = 'entropy')\n\nclf.fit(X_train, y_train)\n\ny_pred = clf.predict(X_test)\n\nfrom sklearn.metrics import accuracy_score\nprint('Accuracy Score on train data: ', accuracy_score(y_true=y_train, y_pred=clf.predict(X_train)))\nprint('Accuracy Score on test data: ', accuracy_score(y_true=y_test, y_pred=y_pred))\n\n# +\nfrom xgboost import XGBClassifier\nxgb = XGBClassifier()\nxgb.fit(X_train,y_train)\nprint(f'Train Score: {xgb.score(X,y)}')\n\ny_pred = xgb.predict(X_test)\nprint(f'Test Accuracy: {accuracy_score(y_test,y_pred)}')\n# -\n\n\n","repo_name":"belsarej/cardiovascular-Diesease-Predcition","sub_path":"Cardiovascular disease prediction.ipynb","file_name":"Cardiovascular disease prediction.ipynb","file_ext":"py","file_size_in_byte":3634,"program_lang":"python","lang":"en","doc_type":"code","stars":0,"dataset":"github-jupyter-script","pt":"44"}
+{"seq_id":"22631985252","text":"# + [markdown] id=\"view-in-github\" colab_type=\"text\"\n# \n\n# + id=\"qET6IUyof18T\"\nimport yfinance as yf\nimport pandas as pd\nimport numpy as np\nimport matplotlib.pyplot as plt\n\n# + colab={\"base_uri\": \"https://localhost:8080/\"} id=\"snxmCYs9gAPV\" outputId=\"655d082a-f535-4a5f-a996-892fadea0007\"\ndatos=yf.download(\"^GSPC\",period=\"1y\",interval=\"1h\")\n\n# + colab={\"base_uri\": \"https://localhost:8080/\"} id=\"zbpIMZ8bgAxu\" outputId=\"aaf2c989-258a-495e-f932-86d8e9829dbb\"\n# Cargar los datos del archivo CSV\ndata = datos\n\n# Calcular los retornos diarios\ndata[\"Returns\"] = data[\"Close\"].pct_change()\n\n# Función para calcular el rendimiento acumulado\ndef calcular_rendimiento(periodo_corta, periodo_larga):\n # Calcular medias móviles\n data[\"SMA_Corta\"] = data[\"Close\"].rolling(window=periodo_corta).mean()\n data[\"SMA_Larga\"] = data[\"Close\"].rolling(window=periodo_larga).mean()\n\n # Generar señales de entrada y salida\n data[\"Signal\"] = 0\n data.loc[data[\"SMA_Corta\"] > data[\"SMA_Larga\"], \"Signal\"] = 1\n data.loc[data[\"SMA_Corta\"] < data[\"SMA_Larga\"], \"Signal\"] = -1\n\n # Calcular retornos acumulados\n data[\"Strategy_Returns\"] = data[\"Signal\"].shift() * data[\"Returns\"]\n data[\"Cumulative_Returns\"] = (1 + data[\"Strategy_Returns\"]).cumprod()\n\n # Retornar rendimiento acumulado final\n return data[\"Cumulative_Returns\"].iloc[-1]\n\n# Parámetros para la optimización\nperiodos_corta = range(10, 101, 10) # Rango de períodos para media móvil corta\nperiodos_larga = range(50, 251, 25) # Rango de períodos para media móvil larga\n\nmejor_rendimiento = float(\"-inf\")\nmejores_periodos = ()\n\n# Bucle para encontrar los períodos que maximizan el rendimiento\nfor periodo_corta in periodos_corta:\n for periodo_larga in periodos_larga:\n rendimiento = calcular_rendimiento(periodo_corta, periodo_larga)\n if rendimiento > mejor_rendimiento:\n mejor_rendimiento = rendimiento\n mejores_periodos = (periodo_corta, periodo_larga)\n\n# Imprimir los mejores períodos encontrados\nprint(\"Mejores períodos: Media Móvil Corta =\", mejores_periodos[0], \", Media Móvil Larga =\", mejores_periodos[1])\n\n\n# + colab={\"base_uri\": \"https://localhost:8080/\", \"height\": 564} id=\"Fd6g0SUsgUEK\" outputId=\"49328ec0-629e-491f-b7ad-99912f0d2bf0\"\n\n\ndata=datos\n\n# Paso 1: Obtener los datos históricos del S&P 500 del último año\ndata = datos # Reemplaza 'datos_sp500.csv' con el nombre de tu archivo de datos\n\n# Paso 2: Análisis de datos y generación de señales\ndata['SMA40'] = data['Close'].rolling(window=40).mean() # Media móvil de 50 días\ndata['SMA150'] = data['Close'].rolling(window=150).mean() # Media móvil de 200 días\n\n# Generar señales de compra y venta basadas en cruces de medias móviles\ndata['Signal'] = np.where(data['SMA40'] > data['SMA150'], 1, -1)\n\n# Paso 3: Definir reglas de entrada y salida\ndata['Position'] = data['Signal'].shift() # Posición al día siguiente para evitar operar en el mismo día de la señal\n\n# Paso 4: Backtesting y evaluación de la estrategia\ndata['Returns'] = np.log(data['Close'] / data['Close'].shift()) # Retornos logarítmicos diarios\ndata['StrategyReturns'] = data['Position'] * data['Returns']\ncumulative_returns = np.exp(data['StrategyReturns'].cumsum()) # Rendimiento acumulado de la estrategia\n\n# Paso 5: Visualización de los resultados\nplt.figure(figsize=(10, 6))\nplt.plot(cumulative_returns)\nplt.title('Rendimiento acumulado de la estrategia')\nplt.xlabel('Fecha')\nplt.ylabel('Rendimiento acumulado')\nplt.grid(True)\nplt.show()\n\n# + id=\"bKRTBlvPgfWU\"\n\n","repo_name":"karloxkronfeld/Finanzas","sub_path":"FomulaMEDIAS.ipynb","file_name":"FomulaMEDIAS.ipynb","file_ext":"py","file_size_in_byte":3762,"program_lang":"python","lang":"en","doc_type":"code","stars":0,"dataset":"github-jupyter-script","pt":"44"}
+{"seq_id":"25742386408","text":"# # importing libraries\n\nimport pandas as pd\nimport numpy as np\nimport matplotlib.pyplot as plt\nimport seaborn as sns\n\n# # loading dataset\n\ndata = pd.read_csv(\"C:\\\\Users\\\\pc\\\\Desktop\\\\csv datasets\\\\survey lung cancer.csv\")\n\ndata.head(10)\n\nsome_data = data.iloc[:,[10,11,12,13,14,15]]\n\ncancer_smoking = data.iloc[:,[2,15]]\n\n# # smoking count\n\ndata.iloc[:,2]\nsmoking = [i.all() for i in data.iloc[:,2].values==2 if i == True]\nnonsmoking = [i.all() for i in data.iloc[:,2].values == 2 if i==False ]\n\nprint(len(smoking))\nprint(len(nonsmoking))\n\n# # visualization smoking\n\nplt.bar((smoking),len(smoking),width = 0.1)\nplt.bar((nonsmoking),len(nonsmoking),color = 'r',width = 0.1)\nplt.title(\"no. of smokers\")\nplt.xlabel(\"smoking or not\")\nplt.ylabel(\"no. of cases\")\nplt.grid()\n\ndata['LUNG_CANCER'].value_counts()\n\n# # smoking vs cancer\n\ncancer_smoking\n\nplt.hist(data['AGE'],bins=20)\n\nsmoking_cancer = []\nsmoking_nocancer = []\nfor i in range(0,309):\n if data.loc[i, \"SMOKING\"]==2 and data.loc[i,\"LUNG_CANCER\"]=='YES':\n smoking_cancer.append(data.loc[i,\"LUNG_CANCER\"])\n elif data.loc[i,\"SMOKING\"]==2 and data.loc[i,\"LUNG_CANCER\"]=='NO':\n smoking_nocancer.append(data.loc[i,\"LUNG_CANCER\"])\n\nnosmoking_cancer = []\nnosmoking_nocancer = []\nfor i in range(0,309):\n if data.loc[i,\"SMOKING\"]==1 and data.loc[i,\"LUNG_CANCER\"]=='YES':\n nosmoking_cancer.append(data.loc[i,\"LUNG_CANCER\"])\n elif data.loc[i,\"SMOKING\"]==1 and data.loc[i,\"LUNG_CANCER\"]=='NO':\n nosmoking_nocancer.append(data.loc[i,\"LUNG_CANCER\"])\n\nlen(nosmoking_cancer)\n\nlen(nosmoking_nocancer)\n\nlen(smoking_cancer)\n\nlen(smoking_nocancer)\n\n# # visualization smoking vs cancer\n\nplt.bar(smoking_cancer,len(smoking_cancer),width = 0.1)\nplt.bar(smoking_nocancer,len(smoking_nocancer),width = 0.1,color = 'r')\nplt.grid()\nplt.title(\"smoking kills?\")\nplt.xlabel(\"lung_cancer\")\nplt.ylabel(\"cases of smoking\")\n\nplt.bar(nosmoking_cancer,len(nosmoking_cancer),width = 0.1)\nplt.bar(nosmoking_nocancer,len(nosmoking_nocancer),width = 0.1,color = 'r')\nplt.grid()\nplt.title(\"smoking kills?\")\nplt.xlabel(\"lung_cancer\")\nplt.ylabel(\"cases of non-smoking\")\n\n# conclusion: the no. of cases where the patient who did not smoke but still had cancer was reletively lower where the smoker had cancer,still the the diffrence is less, it's safe to say someting other than smoking is largly contributing in cause of cancer, also the diffrence in smoker and non smoker who did nothave cancer is also where low. \n\n# # symptom yellow fingers and what it tells\n\nyellowfingers_cancer = []\nyellowfingers_nocancer = []\nfor i in range(0,309):\n if data.loc[i, \"YELLOW_FINGERS\"]==2 and data.loc[i,\"LUNG_CANCER\"]=='YES':\n yellowfingers_cancer.append(data.loc[i,\"LUNG_CANCER\"])\n elif data.loc[i,\"YELLOW_FINGERS\"]==2 and data.loc[i,\"LUNG_CANCER\"]=='NO':\n yellowfingers_nocancer.append(data.loc[i,\"LUNG_CANCER\"])\n\nnoyellowfingers_cancer = []\nnoyellowfingers_nocancer = []\nfor i in range(0,309):\n if data.loc[i, \"YELLOW_FINGERS\"]==1 and data.loc[i,\"LUNG_CANCER\"]=='YES':\n noyellowfingers_cancer.append(data.loc[i,\"LUNG_CANCER\"])\n elif data.loc[i,\"YELLOW_FINGERS\"]==1 and data.loc[i,\"LUNG_CANCER\"]=='NO':\n noyellowfingers_nocancer.append(data.loc[i,\"LUNG_CANCER\"])\n\nlen(yellowfingers_nocancer)\n\nlen(yellowfingers_cancer)\n\nlen(noyellowfingers_nocancer)\n\nlen(noyellowfingers_cancer)\n\n# # visualising yellow finger symptom vs cancer\n\nplt.bar(yellowfingers_cancer,len(yellowfingers_cancer),width = 0.1)\nplt.bar(yellowfingers_nocancer,len(yellowfingers_nocancer),width = 0.1,color = 'r')\nplt.grid()\nplt.title(\"what does symptoms suggest?\")\nplt.xlabel(\"lung_cancer\")\nplt.ylabel(\"cases of yellowfingers\")\n\nplt.bar(noyellowfingers_cancer,len(noyellowfingers_cancer),width = 0.1)\nplt.bar(noyellowfingers_nocancer,len(noyellowfingers_nocancer),width = 0.1,color = 'r')\nplt.grid()\nplt.title(\"what does symptoms suggest?\")\nplt.xlabel(\"lung_cancer\")\nplt.ylabel(\"cases of non-yellowfingers\")\n\n# conclusion: there are over 160 cases where the patient having yellow fingers symptom was diagnosed cancer positive,where as just over 10 cases where he was diagnosed otherwise,and there are over 20 cases where the patient with no yellowfinger symptoms was diagonosed cancer negative, yellow fingers can be seen as positive symptom \n\n# # alcohol vs cancer\n\nalcohol_cancer = []\nalcohol_nocancer = []\nfor i in range(0,309):\n if data.loc[i,\"ALCOHOL CONSUMING\"]==2 and data.loc[i,\"LUNG_CANCER\"]=='YES':\n alcohol_cancer.append(data.loc[i,\"LUNG_CANCER\"])\n elif data.loc[i,\"ALCOHOL CONSUMING\"]==2 and data.loc[i,\"LUNG_CANCER\"]=='NO':\n alcohol_nocancer.append(data.loc[i,\"LUNG_CANCER\"])\n\nlen(alcohol_cancer)\n\nlen(alcohol_nocancer)\n\nnoalcohol_cancer = []\nnoalcohol_nocancer = []\nfor i in range(0,309):\n if data.loc[i,\"ALCOHOL CONSUMING\"]==1 and data.loc[i,\"LUNG_CANCER\"]=='YES':\n noalcohol_cancer.append(data.loc[i,\"LUNG_CANCER\"])\n elif data.loc[i,\"ALCOHOL CONSUMING\"]==1 and data.loc[i,\"LUNG_CANCER\"]=='NO':\n noalcohol_nocancer.append(data.loc[i,\"LUNG_CANCER\"])\n\nlen(noalcohol_cancer)\n\nlen(noalcohol_nocancer)\n\n# # visualization alcohol vs cancer\n\nplt.bar(alcohol_cancer,len(alcohol_cancer),width = 0.1)\nplt.bar(alcohol_nocancer,len(alcohol_nocancer),width = 0.1,color = 'r')\nplt.grid()\nplt.title(\"drinking kills?\")\nplt.xlabel(\"lung_cancer\")\nplt.ylabel(\"cases of drinking\")\n\nplt.bar(noalcohol_cancer,len(noalcohol_cancer),width = 0.1)\nplt.bar(noalcohol_nocancer,len(noalcohol_nocancer),width = 0.1,color = 'r')\nplt.grid()\nplt.title(\"drinking kills?\")\nplt.xlabel(\"lung_cancer\")\nplt.ylabel(\"cases of non-drinking\")\n\n# conclusion: here we see there are over 160 cases where the patient was drinking and was diaganosed cancer positive, but just about 10 cases when the patient was drinking but was diaganosed cancer negative,we can say drinking is more likely to cause lung cancer as compared to smoking, the above statement is supported by the second graph where we can see it has the highest no. of cancer negative patient who did not do alcohol\n\nsome_data.head(10)\n\n# # symptom coughing vs lung cancer\n\ncoughing_cancer = []\ncoughing_nocancer= []\nfor i in range(0,309):\n if data.loc[i,\"COUGHING\"]==2 and data.loc[i,\"LUNG_CANCER\"]=='YES':\n coughing_cancer.append(data.loc[i,\"LUNG_CANCER\"])\n elif data.loc[i,\"COUGHING\"]==2 and data.loc[i,\"LUNG_CANCER\"]== 'NO':\n coughing_nocancer.append(data.loc[i,\"LUNG_CANCER\"])\n\nnocoughing_cancer = []\nnocoughing_nocancer= []\nfor i in range(0,309):\n if data.loc[i,\"COUGHING\"]==1 and data.loc[i,\"LUNG_CANCER\"]=='YES':\n nocoughing_cancer.append(data.loc[i,\"LUNG_CANCER\"])\n elif data.loc[i,\"COUGHING\"]==1 and data.loc[i,\"LUNG_CANCER\"]== 'NO':\n nocoughing_nocancer.append(data.loc[i,\"LUNG_CANCER\"])\n\nprint(f'coughing and cancer= {len(coughing_cancer)}')\nprint(f'coughing but no cancer= {len(coughing_nocancer)}')\nprint(f'no coughing but cancer= {len(nocoughing_cancer)}')\nprint(f'no coughing and no cancer= {len(nocoughing_nocancer)}')\n\n# # visualizing coughing vs cancer\n\nplt.bar(coughing_cancer,len(coughing_cancer),width = 0.1)\nplt.bar(coughing_nocancer,len(coughing_nocancer),width = 0.1,color = 'r')\nplt.grid()\nplt.title(\"what does symptoms suggest?\")\nplt.xlabel(\"lung_cancer\")\nplt.ylabel(\"cases of coughing\")\n\nplt.bar(nocoughing_cancer,len(nocoughing_cancer),width = 0.1)\nplt.bar(nocoughing_nocancer,len(nocoughing_nocancer),width = 0.1,color = 'r')\nplt.grid()\nplt.title(\"what does symptoms suggest?\")\nplt.xlabel(\"lung_cancer\")\nplt.ylabel(\"cases of not coughing\")\n\n# conclusion: there are over 160 cases where the person diagnosed cancer positive was suffering from coughing where as only 10 cases where did not have cancer when he showed coughing as symptom, whereas there are about 30 cases where person who did not show coughing as symptom was also diagnosed cancer negative.so we can say coughing can be assumed evident symptom for cancer. \n\n# # swallowing difficulties vs lung cancer\n\nswallowingdificulties_cancer = []\nswallowingdificulties_nocancer= []\nfor i in range(0,309):\n if data.loc[i,\"SWALLOWING DIFFICULTY\"]==2 and data.loc[i,\"LUNG_CANCER\"]=='YES':\n swallowingdificulties_cancer.append(data.loc[i,\"LUNG_CANCER\"])\n elif data.loc[i,\"SWALLOWING DIFFICULTY\"]==2 and data.loc[i,\"LUNG_CANCER\"]== 'NO':\n swallowingdificulties_nocancer.append(data.loc[i,\"LUNG_CANCER\"])\n\nnoswallowingdificulties_cancer = []\nnoswallowingdificulties_nocancer= []\nfor i in range(0,309):\n if data.loc[i,\"SWALLOWING DIFFICULTY\"]==1 and data.loc[i,\"LUNG_CANCER\"]=='YES':\n noswallowingdificulties_cancer.append(data.loc[i,\"LUNG_CANCER\"])\n elif data.loc[i,\"SWALLOWING DIFFICULTY\"]==1 and data.loc[i,\"LUNG_CANCER\"]== 'NO':\n noswallowingdificulties_nocancer.append(data.loc[i,\"LUNG_CANCER\"])\n\nprint(f'swallowing dificulties and cancer= {len(swallowingdificulties_cancer)}')\nprint(f'swallowing dificulties but no cancer= {len(swallowingdificulties_nocancer)}')\nprint(f'no swallowing dificulties but cancer= {len(noswallowingdificulties_cancer)}')\nprint(f'no sawllowing dificulties and no cancer= {len(noswallowingdificulties_nocancer)}')\n\n# # visualizing swallowing dificulties vs lung cancer\n\nplt.bar(swallowingdificulties_cancer,len(swallowingdificulties_cancer),width = 0.1)\nplt.bar(swallowingdificulties_nocancer,len(swallowingdificulties_nocancer),width = 0.1,color = 'r')\nplt.grid()\nplt.title(\"what does symptoms suggest?\")\nplt.xlabel(\"lung_cancer\")\nplt.ylabel(\"cases of swallowing dificulties\")\n\nplt.bar(noswallowingdificulties_cancer,len(noswallowingdificulties_cancer),width = 0.1)\nplt.bar(noswallowingdificulties_nocancer,len(noswallowingdificulties_nocancer),width = 0.1,color = 'r')\nplt.grid()\nplt.title(\"what does symptoms suggest?\")\nplt.xlabel(\"lung_cancer\")\nplt.ylabel(\"cases of no swallowing dificulties\")\n\n# conclusion:from the graph we can see there are about 140 cases where the patient diagnosed cancer positive had swallowing dificulties as symptom and only 5 cases where the patient having difficulty swallowing was diagnosed cancer negative\n# and also there are more than 30 cases where patient did not have difficulty swallowing and was also diagnosed cancer negative\n# so with we can say difficulties swallowing is one of the major symptoms of lung cancer\n\n# # chest pain vs lung cancer\n\nsome_data.head(10)\n\nchestpain_cancer = []\nchestpain_nocancer= []\nfor i in range(0,309):\n if data.loc[i,\"CHEST PAIN\"]==2 and data.loc[i,\"LUNG_CANCER\"]=='YES':\n chestpain_cancer.append(data.loc[i,\"LUNG_CANCER\"])\n elif data.loc[i,\"CHEST PAIN\"]==2 and data.loc[i,\"LUNG_CANCER\"]== 'NO':\n chestpain_nocancer.append(data.loc[i,\"LUNG_CANCER\"])\n\nnochestpain_cancer = []\nnochestpain_nocancer= []\nfor i in range(0,309):\n if data.loc[i,\"CHEST PAIN\"]==1 and data.loc[i,\"LUNG_CANCER\"]=='YES':\n nochestpain_cancer.append(data.loc[i,\"LUNG_CANCER\"])\n elif data.loc[i,\"CHEST PAIN\"]==1 and data.loc[i,\"LUNG_CANCER\"]== 'NO':\n nochestpain_nocancer.append(data.loc[i,\"LUNG_CANCER\"])\n\nprint(f'chest pain and cancer= {len(chestpain_cancer)}')\nprint(f'chest pain but no cancer= {len(chestpain_nocancer)}')\nprint(f'no chest pain but cancer= {len(nochestpain_cancer)}')\nprint(f'no chest pain and no cancer= {len(nochestpain_nocancer)}')\n\n# # visualizing chest pain vs lung cancer\n\nplt.bar(chestpain_cancer,len(chestpain_cancer),width = 0.1)\nplt.bar(chestpain_nocancer,len(chestpain_nocancer),width = 0.1,color = 'r')\nplt.grid()\nplt.title(\"what does symptoms suggest?\")\nplt.xlabel(\"lung_cancer\")\nplt.ylabel(\"cases of chest pain\")\n\nplt.bar(nochestpain_cancer,len(nochestpain_cancer),width = 0.1)\nplt.bar(nochestpain_nocancer,len(nochestpain_nocancer),width = 0.1,color = 'r')\nplt.grid()\nplt.title(\"what does symptoms suggest?\")\nplt.xlabel(\"lung_cancer\")\nplt.ylabel(\"cases of no chest pain\")\n\nfrom sklearn.preprocessing import LabelEncoder\nle = LabelEncoder()\ndata['GENDER'] = le.fit_transform(data['GENDER'])\ndata['LUNG_CANCER'] = le.fit_transform(data['LUNG_CANCER'])\nx = data.iloc[:,:-1].values\ny= data.iloc[:,-1].values\nfrom sklearn.preprocessing import StandardScaler\nsc =StandardScaler()\nx = sc.fit_transform(x)\n\ndata.head(10)\n\n# # using logistic regression\n\nfrom sklearn.model_selection import train_test_split\nx_train,x_test,y_train,y_actual=train_test_split(x,y,test_size=0.3,random_state=0)\nfrom sklearn.linear_model import LogisticRegression\nlr = LogisticRegression()\nlr.fit(x_train,y_train)\n\nlogistic_pred = lr.predict(x_test)\nfrom sklearn.metrics import confusion_matrix,accuracy_score\nprint(f'confusion matrix: {confusion_matrix(logistic_pred,y_actual)}')\nprint(f'accuracy score: {accuracy_score(logistic_pred,y_actual)}')\n\n# # naive bayes algo.\n\nfrom sklearn.naive_bayes import GaussianNB\ngnb = GaussianNB()\ngnb.fit(x_train,y_train)\n\nnaivebayes_pred= gnb.predict(x_test)\nprint(f'confusion matrix: {confusion_matrix(naivebayes_pred,y_actual)}')\nprint(f'accuracy score: {accuracy_score(naivebayes_pred,y_actual)}')\n\n# # kernel svm\n\nfrom sklearn.svm import SVC\nsvm = SVC(kernel='rbf')\nsvm.fit(x_train,y_train)\n\nsvc_pred = svm.predict(x_test)\nprint(f'confusion matrix: {confusion_matrix(svc_pred,y_actual)}')\nprint(f'accuracy score: {accuracy_score(svc_pred,y_actual)}')\n\nsvm2 = SVC(kernel='linear')\nsvm2.fit(x_train,y_train)\nsvc_linear_pred = svm2.predict(x_test)\nprint(f'confusion matrix: {confusion_matrix(svc_linear_pred,y_actual)}')\nprint(f'accuracy score: {accuracy_score(svc_linear_pred,y_actual)}')\n\n# # k nearest neighbors\n\nfrom sklearn.neighbors import KNeighborsClassifier\nknc = KNeighborsClassifier(n_neighbors=5)\nknc.fit(x_train,y_train)\nknc_pred = knc.predict(x_test)\nprint(f'confusion matrix: {confusion_matrix(knc_pred,y_actual)}')\nprint(f'accuracy score: {accuracy_score(knc_pred,y_actual)}')\n\n# # decision tree clasifier\n\nfrom sklearn.tree import DecisionTreeClassifier\ndtc = DecisionTreeClassifier(criterion='entropy')\ndtc.fit(x_train,y_train)\ndtc_pred = dtc.predict(x_test)\nprint(f'confusion matrix: {confusion_matrix(dtc_pred,y_actual)}')\nprint(f'accuracy score: {accuracy_score(dtc_pred,y_actual)}')\n\n# # random forest classifier\n\nfrom sklearn.ensemble import RandomForestClassifier\nrfc = RandomForestClassifier(n_estimators=5,criterion='entropy')\nrfc.fit(x_train,y_train)\nrfc_pred = rfc.predict(x_test)\nprint(f'confusion matrix: {confusion_matrix(rfc_pred,y_actual)}')\nprint(f'accuracy score: {accuracy_score(rfc_pred,y_actual)}')\n\nlogistic_acc = accuracy_score(logistic_pred,y_actual)\nkernel_pred_linear_acc = accuracy_score(svc_linear_pred,y_actual)\nnaive_acc = accuracy_score(naivebayes_pred,y_actual)\nknearest_acc = accuracy_score(knc_pred,y_actual)\ndecision_acc = accuracy_score(dtc_pred,y_actual)\nrandom_forest_acc = accuracy_score(rfc_pred,y_actual)\n\n# +\nmodels = ['Logistic Regression', 'KNN','naive bayes', 'SVC', 'Decision Tree', 'Random Forest']\nscores = [logistic_acc,knearest_acc,naive_acc,kernel_pred_linear_acc,decision_acc,random_forest_acc]\n\nmodels = pd.DataFrame({'Model' : models, 'Score' : scores})\n\n\nmodels.sort_values(by = 'Score', ascending = False)\n# -\n\n# conclusion: the random forest classification model performed best with 90% accuracy \n","repo_name":"ANUBHAVGithub-code/lung_cancer_prediction_with_EDA_and_ML_models","sub_path":"lung_cancer_prediction_with_EDA_and_ML_models.ipynb","file_name":"lung_cancer_prediction_with_EDA_and_ML_models.ipynb","file_ext":"py","file_size_in_byte":15107,"program_lang":"python","lang":"en","doc_type":"code","stars":0,"dataset":"github-jupyter-script","pt":"40"}
+{"seq_id":"19675664381","text":"# %matplotlib inline\n\nfrom IPython.html.widgets import interact\nimport matplotlib.pyplot as plt\nimport numpy as np\n\n\ndef plot_sine(n,a):\n x = np.arange(0, 20, 0.1);\n y = a*np.sin(x/n)\n plt.plot(x, y)\n\n\n\ninteract(plot_sine, n=(1,10, 0.1), a=(0.2,2, 0.1))\n\n# +\nimport numpy as np\nimport matplotlib.pyplot as plt\nimport matplotlib as mpl\nfrom IPython.html.widgets import interact, FloatSlider, RadioButtons\n\n# mpl.rcParams['figure.max_open_warning'] = 1\n\ndef plot(amplitude, color):\n\n fig, ax = plt.subplots(figsize=(10, 10),\n subplot_kw={'axisbg':'#EEEEEE',\n 'axisbelow':True})\n\n ax.grid(color='w', linewidth=2, linestyle='solid')\n x = np.linspace(0, 10, 1000)\n ax.plot(x, amplitude * np.sin(x), color=color,\n lw=5, alpha=0.4)\n ax.set_xlim(0, 10)\n ax.set_ylim(-1.1, 1.1)\n return fig\n\nres = interact(\n plot,\n amplitude=FloatSlider(min=0.1, max=1.0, step=0.1, value=0.2),\n color=RadioButtons(options=['blue', 'green', 'red'])\n)\n","repo_name":"ricleal/PythonCode","sub_path":"Widgets/sin2.ipynb","file_name":"sin2.ipynb","file_ext":"py","file_size_in_byte":1038,"program_lang":"python","lang":"en","doc_type":"code","stars":0,"dataset":"github-jupyter-script","pt":"40"}
+{"seq_id":"73683520119","text":"# # Analyzing Fandango's movie ratings\n#\n# We will look at the Fandango's movie website to see if their rating system is suspicious, as a previous analysis (made by Walter Hickey) revealed that their ratings were inflated.\n\nimport numpy as np\nimport pandas as pd\nimport matplotlib.pyplot as plt\nimport seaborn as sns\n# %matplotlib inline\n\n# ## Load up files\n#\n# Loading up the two datasets that we will use, an old one with the data before Hickey's analysis, and a newer one with data from after.\n\nold = pd.read_csv('https://raw.githubusercontent.com/fivethirtyeight/data/master/fandango/fandango_score_comparison.csv')\nold.info()\n\nnew = pd.read_csv('https://github.com/mircealex/Movie_ratings_2016_17/raw/master/movie_ratings_16_17.csv')\nnew.info()\n\n# Let's isolate the variables of interest for our analysis:\n\n# +\nfan_old = old[['FILM', 'Fandango_Stars', 'Fandango_Ratingvalue', 'Fandango_votes',\n 'Fandango_Difference']].copy()\nfan_new = new[['movie', 'year', 'fandango']].copy()\n\nfan_old.head(3)\n# -\n\nfan_new.head()\n\n# Our goal is to determine whether there has been any change in Fandango's rating system after Hickey's analysis. The population of interest for our analysis is made of all the movie ratings stored on Fandango's website, regardless of the releasing year.\n#\n# Because we want to find out whether the parameters of this population changed after Hickey's analysis, we're interested in sampling the population at two different periods in time — previous and after Hickey's analysis — so we can compare the two states.\n#\n# The data we're working with was sampled at the moments we want: one sample was taken previous to the analysis, and the other after the analysis. We want to describe the population, so we need to make sure that the samples are representative, otherwise we should expect a large sampling error and, ultimately, wrong conclusions.\n#\n# From Hickey's article and from the README.md of the data set's repository, we can see that he used the following sampling criteria:\n#\n# The movie must have had at least 30 fan ratings on Fandango's website at the time of sampling (Aug. 24, 2015).\n# The movie must have had tickets on sale in 2015.\n# The sampling was clearly not random because not every movie had the same chance to be included in the sample — some movies didn't have a chance at all (like those having under 30 fan ratings or those without tickets on sale in 2015). It's questionable whether this sample is representative of the entire population we're interested to describe. It seems more likely that it isn't, mostly because this sample is subject to temporal trends — e.g. movies in 2015 might have been outstandingly good or bad compared to other years.\n#\n# The sampling conditions for our other sample were (as it can be read in the README.md of the data set's repository):\n#\n# The movie must have been released in 2016 or later.\n# The movie must have had a considerable number of votes and reviews (unclear how many from the README.md or from the data).\n# This second sample is also subject to temporal trends and it's unlikely to be representative of our population of interest.\n#\n# Both these authors had certain research questions in mind when they sampled the data, and they used a set of criteria to get a sample that would fit their questions. Their sampling method is called purposive sampling (or judgmental/selective/subjective sampling). While these samples were good enough for their research, they don't seem too useful for us.\n\n# ## Changing the Goal of our Analysis\n#\n# At this point, we can either collect new data or change our the goal of our analysis. We choose the latter and place some limitations on our initial goal.\n#\n# Instead of trying to determine whether there has been any change in Fandango's rating system after Hickey's analysis, our new goal is to determine whether there's any difference between Fandango's ratings for popular movies in 2015 and Fandango's ratings for popular movies in 2016. This new goal should also be a fairly good proxy for our initial goal.\n#\n# Let's first decide which are the popular movies (we need a criterion), and then see how many of these are available in our two datasets.\n\nfan_old['Fandango_votes'].describe()\n\n# The \"old\" database definitely has a lot of popular movies. The other dataset does not have information on the number of votes. How do we check if the movies in this dataset are popular? One way is to quickly sample a few movies from the dataset and then check manually.\n\nfan_new.sample(10)\n\n# Now we need to isolate the movies from 2015 for the old database, and movies of 2016 for the new database.\n\nfan_old.head()\n\nfan_old['Year'] = fan_old['FILM'].str.extract(r'(?!\\()(\\d{4})(?=\\))')\nfan_old.head()\n\nfan_old['Year'].value_counts()\n\nfan_old_2015 = fan_old[fan_old['Year']=='2015'].copy()\nprint(fan_old_2015.info())\nfan_old_2015.head()\n\nfan_new.head()\n\nfan_new_2016 = fan_new[fan_new['year']==2016].copy()\nfan_new_2016.head()\n\n# Our goal is to determine whether there's any difference between Fandango's ratings for popular movies in 2015 and Fandango's ratings for popular movies in 2016.\n#\n# There are many ways we can go about with our analysis, but let's start simple with making a high-level comparison between the shapes of the distributions of movie ratings for both samples.\n\nplt.style.use('fivethirtyeight')\nfan_old_2015['Fandango_Stars'].plot.kde(label='2015')\nfan_new_2016['fandango'].plot.kde(label='2016')\nplt.legend()\nplt.xlabel('Movie Rating (0-5)', size=12)\nplt.xlim(0,5)\nplt.xticks([0, 0.5, 1.0, 1.5, 2.0, 2.5, 3.0, 3.5, 4.0, 4.5, 5.0])\nplt.title('Fandango Movie Ratings for 2015 and 2016', size=20, y=1.07)\nplt.show()\n\n# Two aspects are striking on the figure above:\n#\n# Both distributions are strongly left skewed.\n# The 2016 distribution is slightly shifted to the left relative to the 2015 distribution.\n# The left skew suggests that movies on Fandango are given mostly high and very high fan ratings. Coupled with the fact that Fandango sells tickets, the high ratings are a bit dubious. It'd be really interesting to investigate this further — ideally in a separate project, since this is quite irrelevant for the current goal of our analysis.\n#\n# The slight left shift of the 2016 distribution is very interesting for our analysis. It shows that ratings were slightly lower in 2016 compared to 2015. This suggests that there was a difference indeed between Fandango's ratings for popular movies in 2015 and Fandango's ratings for popular movies in 2016. We can also see the direction of the difference: the ratings in 2016 were slightly lower compared to 2015.\n\nfan_old_2015['Fandango_Stars'].value_counts(normalize=True).sort_index()*100\n\nfan_new_2016['fandango'].value_counts(normalize=True).sort_index()*100\n\n# Even looking at the frequency tables of ratings in the two datasets, it seems pretty clear that the old data has lots of movies (~38%) with a rating of 4.5. The new dataset is slightly more normally distributed, with 40% of movies having a 4.0 rating.\n#\n# Now we'll compare the two datasets for a few summary statistics.\n\nprint('Means of old and new datasets\\' ratings')\nprint('old:',fan_old_2015['Fandango_Stars'].mean(),\" new:\",fan_new_2016['fandango'].mean())\n\nprint('Medians of old and new datasets\\' ratings')\nprint('old:',fan_old_2015['Fandango_Stars'].median(),\" new:\",fan_new_2016['fandango'].median())\n\nprint('Modes of old and new datasets\\' ratings')\nprint('old:',fan_old_2015['Fandango_Stars'].mode()[0],\" new:\",fan_new_2016['fandango'].mode()[0])\n\n# +\nmean_2015 = fan_old_2015['Fandango_Stars'].mean()\nmean_2016 = fan_new_2016['fandango'].mean()\n\nmedian_2015 = fan_old_2015['Fandango_Stars'].median()\nmedian_2016 = fan_new_2016['fandango'].median()\n\nmode_2015 = fan_old_2015['Fandango_Stars'].mode()[0] # the output of Series.mode() is a bit uncommon\nmode_2016 = fan_new_2016['fandango'].mode()[0]\n\nsummary = pd.DataFrame()\nsummary['2015'] = [mean_2015, median_2015, mode_2015]\nsummary['2016'] = [mean_2016, median_2016, mode_2016]\nsummary.index = ['mean', 'median', 'mode']\nround(summary,2)\n\n# +\nplt.style.use('fivethirtyeight')\nsummary['2015'].plot.bar(color = '#0066FF', align = 'center', label = '2015', width = .25)\nsummary['2016'].plot.bar(color = '#CC0000', align = 'edge', label = '2016', width = .25,\n rot = 0, figsize = (8,5))\n\nplt.title('Comparing summary statistics: 2015 vs 2016', y = 1.07)\nplt.ylim(0,5.5)\nplt.yticks(np.arange(0,5.1,.5))\nplt.ylabel('Stars')\nplt.legend(framealpha = 0, loc = 'upper center')\nplt.show()\n# -\n\n# ## Conclusion\n#\n# Our analysis showed that there's indeed a slight difference between Fandango's ratings for popular movies in 2015 and Fandango's ratings for popular movies in 2016. We also determined that, on average, popular movies released in 2016 were rated lower on Fandango than popular movies released in 2015.\n#\n# We cannot be completely sure what caused the change, but the chances are very high that it was caused by Fandango fixing the biased rating system after Hickey's analysis.\n\n\n","repo_name":"mahikappa/fandango_bias","sub_path":"Fandango movie bias.ipynb","file_name":"Fandango movie bias.ipynb","file_ext":"py","file_size_in_byte":9079,"program_lang":"python","lang":"en","doc_type":"code","stars":0,"dataset":"github-jupyter-script","pt":"40"}
+{"seq_id":"27300980479","text":"# # Continuous gradients in colorectal tumor\n\n# +\nimport sys\nimport os\nfrom collections import defaultdict\nimport pandas as pd\nimport scanpy as sc\nimport numpy as np\nimport matplotlib.pyplot as plt\nfrom glmpca import glmpca\nfrom itertools import combinations\nimport torch\n\nimport sys\nfrom importlib import reload\n\nimport gaston\nfrom gaston import neural_net,cluster_plotting, dp_related, segmented_fit, restrict_spots\nfrom gaston import binning_and_plotting, isodepth_scaling, run_slurm_scripts, parse_adata\nfrom gaston import spatial_gene_classification, plot_cell_types, filter_genes, process_NN_output\n\nimport seaborn as sns\nimport math\n# -\n\n# ## Step 1: Pre-processing\n#\n# Here, GASTON takes as input the output of Space Ranger output. In particular the following files are needed: `filtered_feature_bc_matrix.h5` and `spatial/tissue_positions_list.csv`.\n# Optionally, if one wants to use RGB values as output features, the file `spatial/tissue_hires_image.png`.\n\n# !mkdir -p tumor_tutorial_outputs\n\n# +\ndata_folder='colorectal_tumor_data/'\nuse_RGB=True # set to False if you do not want to use RGB as features\n\ncounts_mat, coords_mat, gene_labels, rgb_mean=parse_adata.get_gaston_input_adata(data_folder, get_rgb=use_RGB, spot_umi_threshold=50)\n\n# save matrices\nnp.save('colorectal_tumor_data/counts_mat.npy', counts_mat)\nnp.save('colorectal_tumor_data/coords_mat.npy', coords_mat)\nnp.save('colorectal_tumor_data/gene_labels.npy', gene_labels)\n# -\n\n# Compute GLM-PCs (we use 5)\n\n# +\nnum_dims=5\npenalty=20 # may need to increase if this is too small\n\nglmpca_res=glmpca.glmpca(counts_mat.T, num_dims, fam=\"poi\", penalty=penalty, verbose=True)\nA = glmpca_res['factors'] # should be of size N x num_dims, where each column is a PC\n\nif use_RGB:\n A=np.hstack((A,rgb_mean)) # attach to RGB mean\nnp.save('colorectal_tumor_data/glmpca.npy', A)\n# -\n\n# visualize top GLM-PCs and RGB mean\nrotated_coords=dp_related.rotate_by_theta(coords_mat, -np.pi/2)\nR=2\nC=4\nfig,axs=plt.subplots(R,C,figsize=(20,10))\nfor r in range(R):\n for c in range(C):\n i=r*C+c\n axs[r,c].scatter(rotated_coords[:,0], rotated_coords[:,1], c=A[:,i],cmap='Reds',s=3)\n if i < num_dims:\n axs[r,c].set_title(f'GLM-PC{i}')\n else:\n axs[r,c].set_title('RGB'[i-num_dims])\n\n# ## Step 2: Train GASTON neural network\n\n# We include how to train the neural network with two options: (1) a command line script and (2) in a notebook. We typically train the neural network 30 different times, each with a different seed, and we use the NN with lowest loss. \n\n# ### Option 1: Slurm (Recommended)\n\n# For option (1), the code below creates 30 different Slurm jobs, one for each initialization. To train the NN for a single initialization run the following command:\n#\n# `gaston -i /path/to/coords.npy -o /path/to/glmpca.npy -d /path/to/output_dir -e 10000 -c 500 -p 20 20 -x 20 20 -z adam -s SEED`\n#\n# for a given `SEED` value (integer)\n#\n# NOTE: please wait for models to finish training before running below code. You can check on their status by running `squeue -u uchitra` (replacing `uchitra` with your username)\n\n# +\n# LOAD DATA generated above\n# path_to_glmpca='colorectal_tumor_data/glmpca.npy'\n# path_to_coords='colorectal_tumor_data/coords_mat.npy'\n\n# To approximately recreate paper figures, use same GLM-PCs from paper \npath_to_glmpca='colorectal_tumor_data/glmpca_from_paper.npy'\npath_to_coords='colorectal_tumor_data/coords_from_paper.npy'\n\n# GASTON NN parameters\nisodepth_arch=[20,20] # architecture (two hidden layers of size 20) for isodepth neural network d(x,y)\nexpression_arch=[20,20] # architecture (two hidden layers of size 20) for 1-D expression function\nepochs = 10000 # number of epochs to train NN\ncheckpoint = 500 # save model after number of epochs = multiple of checkpoint\noptimizer = \"adam\"\nnum_restarts=30 # number of initializations\n\noutput_dir='colorectal_tumor_tutorial_outputs' # folder to save model runs\n\n# REPLACE with your own conda environment name and path\nconda_environment='gaston-package'\npath_to_conda_folder='/n/fs/ragr-data/users/uchitra/miniconda3/bin/activate'\n\nrun_slurm_scripts.train_NN_parallel(path_to_coords, path_to_glmpca, isodepth_arch, expression_arch, \n output_dir, conda_environment, path_to_conda_folder,\n epochs=epochs, checkpoint=checkpoint, \n num_seeds=num_restarts)\n# -\n\n# ### Option 2: train in notebook\n\n# +\npath_to_glmpca='colorectal_tumor_data/glmpca_from_paper.npy'\npath_to_coords='colorectal_tumor_data/coords_from_paper.npy'\n\nA=np.load(path_to_glmpca) # GLM-PCA results used in manuscript\nS=np.load(path_to_coords)\n\n# z-score normalize S and A\nS_torch, A_torch = neural_net.load_rescale_input_data(S,A)\n\n# +\n######################################\n# NEURAL NET PARAMETERS (USER CAN CHANGE)\n# architectures are encoded as list, eg [20,20] means two hidden layers of size 20 hidden neurons\nisodepth_arch=[20,20] # architecture for isodepth neural network d(x,y) : R^2 -> R \nexpression_fn_arch=[20,20] # architecture for 1-D expression function h(w) : R -> R^G\n\nnum_epochs = 10000 # number of epochs to train NN (NOTE: it is sometimes beneficial to train longer)\ncheckpoint = 500 # save model after number of epochs = multiple of checkpoint\nout_dir='colorectal_tumor_tutorial_outputs' # folder to save model runs\noptimizer = \"adam\"\nnum_restarts=30\n\n######################################\n\nseed_list=range(num_restarts)\nfor seed in seed_list:\n print(f'training neural network for seed {seed}')\n out_dir_seed=f\"{out_dir}/rep{seed}\"\n os.makedirs(out_dir_seed, exist_ok=True)\n mod, loss_list = neural_net.train(S_torch, A_torch,\n S_hidden_list=isodepth_arch, A_hidden_list=expression_fn_arch, \n epochs=num_epochs, checkpoint=checkpoint, \n save_dir=out_dir_seed, optim=optimizer, seed=seed, save_final=True)\n# -\n\n# ## Step 3: Process neural network output\n\n# We use the model trained for the paper for reproducibility. If you use the model trained above, then figures will closely match the manuscript --- but not exactly match --- due to PyTorch non-determinism in seeding (see https://github.com/pytorch/pytorch/issues/7068 ). \n\n# +\n# gaston_model, A, S= process_NN_output.process_files('colorectal_tumor_tutorial_outputs') # model trained above\ngaston_model, A, S= process_NN_output.process_files('colorectal_tumor_data/reproduce_tumor') # MATCH PAPER FIGURES\n\n# May need to re-load counts_mat, coords_mat, and gene_labels\ncounts_mat=np.load('colorectal_tumor_data/counts_mat.npy',allow_pickle=True)\ncoords_mat=np.load('colorectal_tumor_data/coords_mat.npy',allow_pickle=True)\ngene_labels=np.load('colorectal_tumor_data/gene_labels.npy',allow_pickle=True)\n# -\n\n# Compute isodepth and GASTON domains \n\n# +\nnum_layers=5\ngaston_isodepth, gaston_labels=dp_related.get_isodepth_labels(gaston_model,A,S,num_layers)\n\n# DATASET-SPECIFIC: so domains are ordered oligodendrocyte to molecular, with increasing isodepth\ngaston_isodepth= np.max(gaston_isodepth) -1 * gaston_isodepth\ngaston_labels=(num_layers-1)-gaston_labels\n# -\n\n# Plot topographic map: isodepth and spatial gradients\n\n# +\nshow_streamlines=True\nrotate = np.radians(-90) # rotate coordinates by -90\narrowsize=2\n\ncluster_plotting.plot_isodepth(gaston_isodepth, S, gaston_model, figsize=(7,6), streamlines=show_streamlines, \n rotate=rotate,arrowsize=arrowsize)\n# -\n\n# Plot GASTON domains\n\ndomain_colors=colors=['plum', 'cadetblue', '#F3D9DC','dodgerblue', '#F44E3F']\ncluster_plotting.plot_clusters(gaston_labels, S, figsize=(6,6), \n colors=domain_colors, s=20, lgd=False, \n show_boundary=True, gaston_isodepth=gaston_isodepth, boundary_lw=5, rotate=rotate)\n\n# ## Continuous gradient analysis\n\n# ### Restrict to domains 1, 2 of tumor\n\n# To isolate the tumor section, we restrict to spots with isodepth lying in a given range. The range of isodepth values will need to be tuned depending on the specific application.\n#\n# In some cases, the tissue geometry cannot be represented with a single isodepth. In this case, we recommend first subsetting your tissue to the specific region of interest (eg from ScanPy clustering), and then running GASTON\n\n# +\n# This is the range we used for reproducing figure papers.\nisodepth_min=4.5\nisodepth_max=6.8\n\ncluster_plotting.plot_clusters_restrict(gaston_labels, S, gaston_isodepth, \n isodepth_min=isodepth_min, isodepth_max=isodepth_max, figsize=(6,6), \n colors=domain_colors, s=20, lgd=False, rotate=rotate)\n# -\n\n# Once you have a range that looks reasonable, then we restrict the isodepth, domain labels, coords matrix, and counts matrix to only be for the green spots\n\n# +\n# Optional: adjust isodepth for physical distance\nadjust_physical=True\nscale_factor=100 # since distance of 1 = 100 microns for 10x Visium\n\n# Optional: plot isodepth for green spots\nplot_isodepth=True\n\n# plotting parameters\nshow_streamlines=True\nrotate=np.radians(-90)\narrowsize=1\n\n\ncounts_mat_restrict, coords_mat_restrict, gaston_isodepth_restrict, gaston_labels_restrict=restrict_spots.restrict_spots(\n counts_mat, coords_mat, S, gaston_isodepth, gaston_labels, \n isodepth_min=isodepth_min, isodepth_max=isodepth_max, \n adjust_physical=adjust_physical, scale_factor=scale_factor,\n plot_isodepth=plot_isodepth, show_streamlines=show_streamlines, \n gaston_model=gaston_model, rotate=rotate, figsize=(6,3), \n arrowsize=arrowsize)\n\n# -\n\n# ### Compute piecewise linear fits\n\n# We restrict to genes with at least 1000 total UMIs, that are not mitochondrial/ribosomal\n\numi_thresh = 1000\nidx_kept, gene_labels_idx=filter_genes.filter_genes(counts_mat, gene_labels, \n umi_threshold=umi_thresh, \n exclude_prefix=['MT-', 'RPL', 'RPS'])\n\n# Compute piecewise linear fit over restricted spots from above\n\n# compute piecewise linear fit for restricted spots\npw_fit_dict=segmented_fit.pw_linear_fit(counts_mat_restrict, gaston_labels_restrict, gaston_isodepth_restrict,\n None, [], idx_kept=idx_kept, umi_threshold=umi_thresh, isodepth_mult_factor=0.01,)\n# for plotting\nbinning_output=binning_and_plotting.bin_data(counts_mat2, gaston_labels2, gaston_isodepth2, \n None, gene_labels, idx_kept=idx_kept, num_bins=15, umi_threshold=umi_thresh)\n\n# Find discontinuous and continuous genes\n\n# +\ndomain_colors=['dodgerblue', '#F44E3F']\n\ndiscont_genes_layer=spatial_gene_classification.get_discont_genes(pw_fit_dict, binning_output,q=0.95)\ncont_genes_layer=spatial_gene_classification.get_cont_genes(pw_fit_dict, binning_output,q=0.8)\n# -\n\n# Plot gene COX7B from manuscript. This is a Type I gene, with only intratumoral variation as shown below (continuous gradient only in domain 1, ie tumor)\n\n# +\ngene_name='COX7B'\nprint(f'gene {gene_name}: discontinuous jump after domain(s) {discont_genes_layer[gene_name]}') \nprint(f'gene {gene_name}: continuous gradient in domain(s) {cont_genes_layer[gene_name]}')\n\n# display log CPM (if you want to do CP500, set offset=500)\noffset=10**6\n\nbinning_and_plotting.plot_gene_pwlinear(gene_name, pw_fit_dict, gaston_labels2, gaston_isodepth2, \n binning_output, cell_type_list=None, pt_size=50, colors=domain_colors, \n linear_fit=True, ticksize=15, figsize=(4,2.5), offset=offset, lw=3,\n domain_boundary_plotting=True)\n# -\n\n# Plot gene SCD from manuscript. This is a Type I gene, with intrastromal and intratumoral variation (continuous gradient in domains 0, 1) but no discontinuity\n\n# +\ngene_name='SCD'\nprint(f'gene {gene_name}: discontinuous after domain(s) {discont_genes_layer[gene_name]}') \nprint(f'gene {gene_name}: continuous in domain(s) {cont_genes_layer[gene_name]}')\n\n# display log CPM (if you want to do CP500, set offset=500)\noffset=10**6\n\nbinning_and_plotting.plot_gene_pwlinear(gene_name, pw_fit_dict, gaston_labels2, gaston_isodepth2, \n binning_output, cell_type_list=None, pt_size=50, colors=domain_colors, \n linear_fit=True, ticksize=15, figsize=(4,2.5), offset=offset, lw=3,\n domain_boundary_plotting=True)\n# -\n\n# Plot gene ACTA2 from manuscript. This is a Type II gene, with both intrastromal variation (gradient in domain 0) and discontinuity\n\n# +\ngene_name='ACTA2'\nprint(f'gene {gene_name}: discontinuous after domain(s) {discont_genes_layer[gene_name]}') \nprint(f'gene {gene_name}: continuous in domain(s) {cont_genes_layer[gene_name]}')\n\n# display log CPM (if you want to do CP500, set offset=500)\noffset=10**6\n\nbinning_and_plotting.plot_gene_pwlinear(gene_name, pw_fit_dict, gaston_labels2, gaston_isodepth2, \n binning_output, cell_type_list=None, pt_size=50, colors=domain_colors, \n linear_fit=True, ticksize=15, figsize=(4,2.5), offset=offset, lw=3,\n domain_boundary_plotting=True)\nplt.ylim((3,5.6))\n# -\n\n# Plot gene TAGLN from manuscript. This is a Type II gene, with both intrastromal variation (gradient in domain 0) and discontinuity\n\n# +\ngene_name='TAGLN'\nprint(f'gene {gene_name}: discontinuous after domain(s) {discont_genes_layer[gene_name]}') \nprint(f'gene {gene_name}: continuous in domain(s) {cont_genes_layer[gene_name]}')\n\n# display log CPM (if you want to do CP500, set offset=500)\noffset=10**6\n\nbinning_and_plotting.plot_gene_pwlinear(gene_name, pw_fit_dict, gaston_labels2, gaston_isodepth2, \n binning_output, cell_type_list=None, pt_size=50, colors=domain_colors, \n linear_fit=True, ticksize=15, figsize=(4,2.5), offset=offset, lw=3,\n domain_boundary_plotting=True)\nplt.ylim((3,6.2))\n# -\n\n# Plot gene COL1A2 from manuscript. This is a Type II gene, with intrastromal/intratumoral variation and discontinuity\n\n# +\ngene_name='COL1A2'\nprint(f'gene {gene_name}: discontinuous after domain(s) {discont_genes_layer[gene_name]}') \nprint(f'gene {gene_name}: continuous in domain(s) {cont_genes_layer[gene_name]}')\n\n# display log CPM (if you want to do CP500, set offset=500)\noffset=10**6\n\nbinning_and_plotting.plot_gene_pwlinear(gene_name, pw_fit_dict, gaston_labels2, gaston_isodepth2, \n binning_output, cell_type_list=None, pt_size=50, colors=domain_colors, \n linear_fit=True, ticksize=15, figsize=(4,2.5), offset=offset, lw=3,\n domain_boundary_plotting=True)\nplt.ylim((4,6.5))\n# -\n\n# Plot gene THBS1 from manuscript. This is a Type III gene, with intrastromal variation\n\n# +\ngene_name='THBS1'\nprint(f'gene {gene_name}: discontinuous after domain(s) {discont_genes_layer[gene_name]}') \nprint(f'gene {gene_name}: continuous in domain(s) {cont_genes_layer[gene_name]}')\n\n# display log CPM (if you want to do CP500, set offset=500)\noffset=10**6\n\nbinning_and_plotting.plot_gene_pwlinear(gene_name, pw_fit_dict, gaston_labels2, gaston_isodepth2, \n binning_output, cell_type_list=None, pt_size=50, colors=domain_colors, \n linear_fit=True, ticksize=15, figsize=(4,2.5), offset=offset, lw=3,\n domain_boundary_plotting=True)\nplt.ylim((3,6))\n# -\n\n\n","repo_name":"raphael-group/GASTON","sub_path":"docs/notebooks/tutorials/tumor_tutorial.ipynb","file_name":"tumor_tutorial.ipynb","file_ext":"py","file_size_in_byte":16035,"program_lang":"python","lang":"en","doc_type":"code","stars":1,"dataset":"github-jupyter-script","pt":"40"}
+{"seq_id":"15810890271","text":"# + id=\"jRi3-5Kd8pkv\" colab_type=\"code\" outputId=\"d1a53b11-0722-4a05-c4b0-d5476dae8f6f\" colab={\"base_uri\": \"https://localhost:8080/\", \"height\": 423}\n#SuperVised Learning\n #1. Classification\n #2. Regression\nimport pandas as pd\niris = pd.read_csv('Iris.csv')\niris\n\n# + id=\"t2C3FLiSBzXt\" colab_type=\"code\" outputId=\"138b114e-5484-4411-894a-23985018087b\" colab={\"base_uri\": \"https://localhost:8080/\", \"height\": 147}\niris.isnull().sum()\n\n# + id=\"ycB7U_PcAeNT\" colab_type=\"code\" outputId=\"76a79ce0-5811-4890-8f9b-680b3e3e09f2\" colab={\"base_uri\": \"https://localhost:8080/\", \"height\": 36}\niris['Species'].unique()\n\n# + id=\"2Woi1S2UBniw\" colab_type=\"code\" colab={}\n#Encoding of labels\nY = iris['Species'].map({'Iris-setosa':0, \n 'Iris-versicolor':1, \n 'Iris-virginica':2 })\n\n# + id=\"2gqFSL5ECY2L\" colab_type=\"code\" outputId=\"8814b5d8-383c-4ef4-d808-91ca78170ae3\" colab={\"base_uri\": \"https://localhost:8080/\", \"height\": 423}\n#Features or X or independent variables\nX = iris.drop(['Id','Species'],axis=1) #axis=1 as we are selecting columns\nX\n\n# + id=\"mlm57mguDKsq\" colab_type=\"code\" colab={}\n#Train your model\nfrom sklearn.neighbors import KNeighborsClassifier\nkmodel = KNeighborsClassifier(n_neighbors=3)\n\n# + id=\"xta_XP2QDd9R\" colab_type=\"code\" colab={}\n#Validate your model\n#what is the accuracy of your model ?\n#Training data set\n#Validation data set/ testing data set\n#xtrain-120 , xtest-30\n#ytrain-120 , ytest-30\nfrom sklearn.model_selection import train_test_split\nxtrain,xtest,ytrain,ytest = train_test_split(X,Y,test_size=0.50)\n\n# + id=\"XhzBTWiJL151\" colab_type=\"code\" outputId=\"d4a35682-edf3-45c2-eb69-3bf890eeeb8f\" colab={\"base_uri\": \"https://localhost:8080/\", \"height\": 73}\nkmodel.fit(xtrain,ytrain)\n\n# + id=\"6bxLcWwoL8Jm\" colab_type=\"code\" outputId=\"a2d70df3-434f-4554-fc55-baecf36c47fa\" colab={\"base_uri\": \"https://localhost:8080/\", \"height\": 91}\nypred = kmodel.predict(xtest)\nypred\n\n# + id=\"F8H3aKkHLYlT\" colab_type=\"code\" outputId=\"ef2ad73a-cb86-4797-c2be-bb975391f839\" colab={\"base_uri\": \"https://localhost:8080/\", \"height\": 36}\n(ytest == ypred).sum()/len(xtest)\n\n# + id=\"GCICnl6IMqTB\" colab_type=\"code\" outputId=\"fa3a1a97-a86c-4ff1-d327-d944a7713500\" colab={\"base_uri\": \"https://localhost:8080/\", \"height\": 36}\nkmodel.score(xtest,ytest)\n\n# + id=\"TM-tc1VTDa-Q\" colab_type=\"code\" outputId=\"dd28750d-717d-42fc-e4e3-a255de8f4399\" colab={\"base_uri\": \"https://localhost:8080/\", \"height\": 36}\n#Predict the output\n#.predict()\nkmodel.predict([[3.2,2.1,8.9,6.8],[2.1,3.2,3.3,2.1]])\n\n# + id=\"EXWEx7sUJdDb\" colab_type=\"code\" outputId=\"0d6669a1-c5b4-425d-d3a3-dafba85633e5\" colab={\"base_uri\": \"https://localhost:8080/\", \"height\": 36}\n#How to decide the value of k\n# How to tune a a parameter in training\nprint('hello')\n\n# + id=\"GIn3gsb9N4p_\" colab_type=\"code\" outputId=\"1b09dc88-860f-4d84-8d3d-fafaf657657d\" colab={\"base_uri\": \"https://localhost:8080/\", \"height\": 36}\nlen(xtrain)\n\n# + id=\"zrSslmlnPwN7\" colab_type=\"code\" colab={}\n#Parameter tuning , k\n#only be decided by error monitoring\naccuracy = []\nfor i in range(1,10):\n km = KNeighborsClassifier(n_neighbors=i)\n km.fit(xtrain,ytrain)\n s = km.score(xtest,ytest)\n accuracy.append(s)\n\n# + id=\"kTRgaHi4P6vL\" colab_type=\"code\" outputId=\"c2aca6a2-2341-45ab-bccb-6cc37a5bcf1d\" colab={\"base_uri\": \"https://localhost:8080/\", \"height\": 265}\naccuracy\nimport numpy as np\na=np.arange(1,10)\nb=accuracy\nimport matplotlib.pyplot as plt\nplt.plot(a,b)\nplt.show()\n\n# + id=\"CpnTZBIZRo09\" colab_type=\"code\" colab={}\n#High Accuracy\n#4 - Features - sl,sw,pl,pw\n#any 2 feature\nx1 = X['SepalLengthCm']\nx2 = X['SepalWidthCm']\nx3 = X['PetalLengthCm']\nx4 = X['PetalWidthCm']\n\n# + id=\"2TvGImWGT-Do\" colab_type=\"code\" outputId=\"810d91f0-654b-4ec5-a25b-cc0a2709de75\" colab={\"base_uri\": \"https://localhost:8080/\", \"height\": 239}\nY\n\n# + id=\"xVtqTobCT_PT\" colab_type=\"code\" outputId=\"26193bb2-9e1b-4747-ef51-7caeb9e6077c\" colab={\"base_uri\": \"https://localhost:8080/\", \"height\": 268}\nplt.scatter(x1,x2,c=Y)\nplt.show()\n\n# + id=\"2Vl-JnIBUHcT\" colab_type=\"code\" outputId=\"1ece32f4-0365-47ac-a19a-e0f72a6879a5\" colab={\"base_uri\": \"https://localhost:8080/\", \"height\": 265}\nplt.scatter(x3,x4,c=Y)\nplt.show()\n\n# + id=\"xzCae-mNU9_Q\" colab_type=\"code\" outputId=\"2d67f9c3-83a3-44ef-b862-638a3ba6efc8\" colab={\"base_uri\": \"https://localhost:8080/\", \"height\": 265}\nplt.scatter(x1,x3,c=Y)\nplt.show()\n\n# + id=\"lbyNE9NWVbYf\" colab_type=\"code\" colab={}\nX['label'] = Y\n\n# + id=\"TYiN6Q2SWg_s\" colab_type=\"code\" outputId=\"d584cc75-a483-458e-e076-9152fc23fea9\" colab={\"base_uri\": \"https://localhost:8080/\", \"height\": 423}\nX\n\n# + id=\"uap8qU9LWpbC\" colab_type=\"code\" outputId=\"a15ff43a-cfe2-4e90-97a9-00a975ecf98b\" colab={\"base_uri\": \"https://localhost:8080/\", \"height\": 206}\nX.corr()\n\n# + id=\"xOF8ylegWtNg\" colab_type=\"code\" outputId=\"1fcdc00c-9ca4-42bf-f4b9-e2a3c70aa4d0\" colab={\"base_uri\": \"https://localhost:8080/\", \"height\": 36}\nkmodel.predict([[3.2,1.2,0.9,8.9],\n [3.2,1.2,0.9,0.9]])\n\n# + id=\"RFd8npeiaazc\" colab_type=\"code\" colab={}\nname = np.array([['setosa'],\n ['versicolor'],\n ['virginica']])\n\n# + id=\"lze8uRCravxl\" colab_type=\"code\" outputId=\"b8517074-3118-4b92-9513-23432d9c7c33\" colab={\"base_uri\": \"https://localhost:8080/\", \"height\": 54}\nname[kmodel.predict([[3.2,1.2,0.9,8.9],[3.2,1.2,0.9,0.9]])]\n\n# + id=\"IQ5ynvWva1tq\" colab_type=\"code\" colab={}\n\n","repo_name":"arnav8/Innovians-Tech","sub_path":"Classification.ipynb","file_name":"Classification.ipynb","file_ext":"py","file_size_in_byte":5354,"program_lang":"python","lang":"en","doc_type":"code","stars":0,"dataset":"github-jupyter-script","pt":"40"}
+{"seq_id":"41308203138","text":"import elasticsearch\nfrom elasticsearch import Elasticsearch \n\nimport tweepy\nfrom tweepy import Stream\nfrom tweepy.streaming import StreamListener\nfrom tweepy import OAuthHandler\nimport time\nimport json\n\nc_key = ''\nsecret_c_key = ''\naccess_token = ''\nsecret_access_token = ''\n\nimport logging\nlogging.getLogger(\"elasticsearch\").setLevel(logging.CRITICAL)\nlogging.basicConfig(level=logging.CRITICAL)\n\nes = Elasticsearch(HOST=\"http://localhost\", PORT=9200) \nelast = Elasticsearch() \n\n\n# +\nclass Listener(StreamListener):\n def __init__(self, number):\n self.counter = 0\n self.number = number\n\n def on_data(self, data):\n dict_data = json.loads(data)\n print (dict_data[\"text\"])\n elast.index(index=\"twitter\",\n doc_type=\"stream\",\n body={\"author\": dict_data[\"user\"][\"screen_name\"],\n \"date\": dict_data[\"created_at\"],\n \"message\": dict_data[\"text\"]})\n self.counter +=1\n if self.counter == self.number:\n return False\n \n return True\n \n def on_error(self, status):\n print (status)\n \nif __name__ == '__main__':\n\n Tweet_Listening = Listener(number = 15) \n Auth = OAuthHandler(c_key, secret_c_key)\n Auth.set_access_token(access_token, secret_access_token)\n Streaming = Stream(Auth, Tweet_Listening)\n Streaming.filter(track=['pizza', 'hamburger'])\n# -\n\n\n\n\n","repo_name":"leopards17/MiniProj3","sub_path":"MiniProj3-Copy1.ipynb","file_name":"MiniProj3-Copy1.ipynb","file_ext":"py","file_size_in_byte":1462,"program_lang":"python","lang":"en","doc_type":"code","stars":0,"dataset":"github-jupyter-script","pt":"42"}
+{"seq_id":"12613386724","text":"import pandas as pd\n\ndf = pd.DataFrame({\n 'Color': ['Red','Blue', 'Green', 'Green', 'Red', 'Blue']\n })\n\ndf\n\nfrom sklearn.preprocessing import LabelEncoder\n\nlabel_encoder = LabelEncoder()\n\nlabel_encoded = pd.DataFrame(label_encoder.fit_transform(df[['Color']]), columns=[['Encoded_Color']])\n\nlabel_encoded\n\npd.concat((df,label_encoded),axis=1)\n","repo_name":"AbhishekWaghchaure/PW_Skills-Data-Science-Masters-course-complete-","sub_path":"March_21_LabelEncoder.ipynb","file_name":"March_21_LabelEncoder.ipynb","file_ext":"py","file_size_in_byte":349,"program_lang":"python","lang":"en","doc_type":"code","stars":0,"dataset":"github-jupyter-script","pt":"42"}
+{"seq_id":"7037043743","text":"# # Outlier Detection of 4 Digit MWC 4 Players using Tournament Scores using Box-Cox Transformation and Average Percentile Model\n\n# ## Introduction\n#\n# This notebook is the explanation and experimentation on an Alternative of Local Outlier Factor with a simpler model, which is an Average Percentile Model.\n\n# ## Logit Transformation\n\n# **Logit Transformation** is a terminology to transform the score which are in range $[0, 1000000]$ into $(-\\infty, \\infty)$ using the formula\n\n# $$\n# \\text{logit}(x) = \\log(\\frac{x}{1 - x})\n# $$\n\n# Where\n# $$\n# x = \\frac{\\text{score}}{1000000}\n# $$\n\n# This will also theoretically widen the gap between small difference of scores when the scores are higher. For example the difference between $\\text{logit}(0.99) - \\text{logit}(0.985)$ is less than $\\text{logit}(0.995) - \\text{logit}(0.99)$. Which matches the environment in the osu!mania tournament scene since the chance that the score gaps are low in earlier game is higher than the late game.\n\n# ## Box-Cox Transformation\n\n# **Box-Cox Transformation** can be written as\n\n# $$\n# \\text{boxcox}(x) = \\begin{cases}\n# \\frac{x^\\lambda - 1}{\\lambda} & \\lambda \\neq 0\\\\\n# \\log(x) & \\lambda = 0\n# \\end{cases}\n# $$\n\n# Theoretically, the following transformation will make the data more Normally Distributed. We can find the $\\lambda$ using **Maximum Likelihood** which will try to find the $\\lambda$ that maximize how \"Normally Distributed\" the data is.\n#\n# More information and experimentations can be found in [boxcox.ipynb](https://github.com/HowToProgramming/4dm4analysis/blob/main/boxcox.ipynb)\n\n# ## Parametric Average Percentile Model\n\n# **Parametric Average Percentile Model** is based on the idea of how to measure the proportion of the population that a given player has surprassed. The data in this case is multi-dimensional data, that means there are multiple maps in the calculation process of this model. First we standardize the data of each map, then we calculate the **Cumulative Distribution Function** of **Normal Distribution** or any given parametric distribution, we will call that a percentile. After that we average the percentile and get the final result of a given player.\n\n# ### Adjustments\n\n# #### Applying Box-Cox Transformation\n\n# As mentioned in the previous section about Box-Cox Transformation, we transform the Logit-Transformed data using Box-Cox transformation in order to use them in the **Parametric Average Percentile Model**. This will maximize the efficiency of the model, however there is a case where $\\lambda < 0$ which we should aware of because the higher the score, the lower the transformed value is, we then use another adjustment\n\n# $$\n# \\text{sign}(\\lambda) = \\begin{cases}\n# 1 & \\lambda \\geq 0\\\\\n# -1 & \\text{otherwise}\n# \\end{cases}\n# $$\n\n# We multiply them with our standardized value and then use that value to calculate the **Cumulative Distribution Function** in order to catch that issue.\n\n# #### Applying Maps Played into the consideration\n\n# After we got the output from our Model, we then find another factor which the model didn't catch which is the \"Maps Played\". The important behaviour of outperformed players is the fact that they usually are main players of most of the maps. So we try to include this feature by multiplying the outlier value that we got from model with the log of maps played. Which is the following formula:\n\n# $$\n# \\text{Adjusted Outlier Score} = \\text{Outlier Value from Avg. Percentile} \\cdot \\log(\\# \\text{Maps Played} + 1)\n# $$\n\n# We then use this value to detect the outlier or outperformed players.\n\n# #### Applying Participation Rate into Consideration\n\n# Another way to evaluate whether a player should be skillbanned would be to consider the participation of a given player in a team. This can be thought as how much a player \"sandbags\" the team. The way to get the participation rate is to divide a player's score with a maximum team score in a given map.\n\n# But how do we prioritize which map should be considered more or less ? This is an important process in skillbanning because if we equally weighted every map, the model might not be reliable due to early game being equal to late game, so we do the following procedure:\n\n# - We obtain the participation rate using the method from a recent paragraph\n# - We calculate the \"probability of elimination\", means how many teams have been eliminated in a round that a given map is placed.\n# - We then combine two values together using multiplication and averaging for each round, then perform the summation to get a participation value.\n\n# The overall formula will be\n\n# $$\n# \\text{Map Participation Rate} = \\frac{\\text{Player Score}}{\\text{Team Score}} \\cdot (\\text{Probability of Elimination} + \\epsilon)\n# $$\n\n# $$\n# \\text{Overall Participation Rate} = \\sum_{\\text{round} \\in \\text{rounds}} \\text{Avg} (\\text{Map Participation Rate})\n# $$\n\n# Then we combine them into outlier values using this formula:\n\n# $$\n# \\text{Adjusted Outlier Values} = \\left(1 + p\\right)x\n# $$\n#\n# Where $p$ is **Overall Participation Rate** and - HowToPlayLN, 2022, while thinking how to write this section
\n#\n# But the more important question is, how do we define outperforming ? Given a player, an easy idea is to calculate how many players a given player has surpressed in a certain map. To be precise, we try to compare the player with the entire population of players who are eligible to participate in a tournament. In order to compare (approximate) directly, we need to use the statistical magic called \"Normal Distribution\".which there is no turning back if we commit it (sry overused joke)
\n#\n# But we cannot use the Normal Distribution directly, from the raw score, we need to use some transformation to extract some features and normalize them. First, we use the \"logit transformation\" to seperate the stacked performances of players in the tournament, for example, we can see more distinction between 990k and 995k score than 985k and 990k, which matches the environment of how osu!mania tournaments usually works. Next, we use the \"Box-Cox Transformation\" in order to make data suitable for the statistical magic.\n#\n# We then input the transformed data into the statistical magic that allows us to approximate how much the player has surprassed the population. However, we do this for every single map in a tournament and use all sample players data to determine the approximated value. Then we average the values for all maps a player played in a tournament.\n#\n# Finally, the last feature comes from the nature of the behaviour of outperforming players in a tournament, especially a country team tournament with the format of 3v3 and 6 players for each country. They usually are the first or second player of the maps. With this information, we can add the \"maps played\" feature into the consideration. We do this by taking a logarithm function to maps played to reduce the difference between players who plays in the late game and focus more on them. We then multiply that value to the previous feature, then we get the numbers that will determine whether the player is outperforming.\n\n# ## Code Documentation\n\n# +\nimport os\n\nos.chdir(\"..\")\n# -\n\n# ### Importing Necessary Modules\n#\n# We will use `pandas` for data management, `matplotlib.pyplot` for data visualization and `numpy` to deal with math stuff.\n\nimport pandas as pd\nimport matplotlib.pyplot as plt\nimport numpy as np\n\n# For Local Packages, we will use `Dataset` and `get_table_players` to transform dataset into the \n\nfrom utils import Dataset\nfrom utils.outlierdetectionmodel import BoxCoxParametricAveragePercent\nfrom utils.dftransformer import get_table_players\n\n_4dm4_dataset = Dataset('datasets/4dm4.db')\n\n_4dm4_data = _4dm4_dataset.select('scores', columns=['player_name', 'round', 'beatmap_type', 'beatmap_tag', 'score_logit'], where={\n 'beatmap_type': ['LN', 'RC', 'HB'],\n})\n\n_4dm4_data.head()\n\nplayer_table = get_table_players(_4dm4_data)\n\nplayer_table\n\nplayer_table[player_table <= 0] = np.nan\n\n# +\nmodel = BoxCoxParametricAveragePercent()\n\nmodel.fit(player_table.values)\n\n# +\ndf = pd.DataFrame(index=player_table.index)\n\ndf['outlier_values'] = model.predict(player_table.values)\n# -\n\ndf.sort_values(by='outlier_values', ascending=False).head(16)\n\ndf['n_maps_played'] = np.sum(pd.notna(player_table.values), axis=1)\n\ndf\n\ndf['adjusted_ol_values_1'] = df['outlier_values'] * np.log(df['n_maps_played'] + 1)\n\ndf.sort_values(by='adjusted_ol_values_1', ascending=False).head(15)\n\nplt.hist(df['outlier_values'])\nplt.show()\n\nplt.hist(df['adjusted_ol_values_1'])\nplt.show()\n\nparticipation_rate_dataset = _4dm4_dataset.query(\"\"\"select scores.player_name, scores.round, scores.beatmap_type, \nscores.beatmap_tag, scores.score as player_score, team_scores.score as team_score\nfrom scores left join player_data on scores.player_name = player_data.player_name\nleft join team_data on team_data.country_code = player_data.country_code\nleft join team_scores on team_data.country_name = team_scores.country_name and\nscores.beatmap_type = team_scores.beatmap_type and scores.round = team_scores.round and scores.beatmap_tag = team_scores.beatmap_tag\nwhere scores.beatmap_type != \\\"SV\\\"\"\"\")\n\nparticipation_rate_dataset\n\nparticipation_rate_dataset['participation_rate'] = participation_rate_dataset['player_score'] / participation_rate_dataset['team_score']\n\nparticipation_rate_dataset\n\nparticipation_rate_dataset = participation_rate_dataset[['player_name', 'round', 'participation_rate']]\n\naverage_participation_rate = participation_rate_dataset.groupby(['player_name', 'round']).mean()\n\naverage_participation_rate\n\n# +\nteam_survived = _4dm4_dataset.query(\"\"\"\nSELECT last_round, count(country_name) from team_data where last_round != \"\" group by last_round \n\"\"\")\n\nround_order = [\"RO16\", \"QF\", \"SF\", \"F\", \"GF\"]\n# -\n\nteam_survived\n\nteam_survived.index = team_survived['last_round']\n\nteam_survived = team_survived.loc[round_order]\n\nteam_survived\n\nteam_survived['country_eliminated'] = team_survived['count(country_name)'].cumsum()\n\nteam_survived\n\n# +\nteam_survived['p_eliminated'] = team_survived['country_eliminated'] / 32\n\nteam_survived\n# -\n\naverage_participation_rate['multipiler'] = average_participation_rate.index.map(lambda x: team_survived[team_survived['last_round'] == x[1]]['p_eliminated'].values + 0.1 or [0.1])\n\naverage_participation_rate\n\naverage_participation_rate['multipiler'] = average_participation_rate['multipiler'].apply(lambda x: x[0])\n\naverage_participation_rate['value'] = average_participation_rate['participation_rate'] * average_participation_rate['multipiler']\n\naverage_participation_rate\n\naverage_participation_rate = average_participation_rate[['value']]\n\naverage_participation_rate = average_participation_rate.groupby(average_participation_rate.index.map(lambda x: x[0])).sum()\n\ndf\n\ndf['participation_value'] = average_participation_rate['value']\n\ndf\n\ndf['adjusted_ol_values_2'] = df['outlier_values'] + df['outlier_values'] * df['participation_value']\n\ndf.sort_values('adjusted_ol_values_2', ascending=False).head(10)\n\n# ---\n","repo_name":"4digitmwc/4dmanalysis","sub_path":"4dm4analysis/averagepercent_detection.ipynb","file_name":"averagepercent_detection.ipynb","file_ext":"py","file_size_in_byte":11301,"program_lang":"python","lang":"en","doc_type":"code","stars":0,"dataset":"github-jupyter-script","pt":"42"} +{"seq_id":"34828445057","text":"# + [markdown] id=\"_GZakiZT-ad8\"\n# **Applying Fractal Clustering o Human Activity recognition data set.**\n\n# + id=\"u7SqTBl9iTN7\"\nimport pandas as pd \nimport numpy as np \nimport matplotlib.pyplot as plt \nimport seaborn as seabornInstance \nfrom sklearn.model_selection import train_test_split \nfrom sklearn.linear_model import LinearRegression\nfrom sklearn import metrics\n# %matplotlib inline\n\n# + colab={\"base_uri\": \"https://localhost:8080/\"} id=\"zGu2rMOYiqC0\" outputId=\"8e51de61-7859-4b46-b8e3-b7128e8adc5c\"\nfrom google.colab import drive\ndrive.mount('/content/drive')\n\n# + id=\"mMoCRlzgixPv\"\nPath = 'drive/My Drive/ML Assignments/dataset'\n\n# + colab={\"base_uri\": \"https://localhost:8080/\", \"height\": 287} id=\"ERiNPLweiyqy\" outputId=\"d877a2ea-e3df-4d6e-a078-e937c1478f9b\"\ndf = pd.read_csv(Path+'/train.csv')\ndf.head()\n\n# + colab={\"base_uri\": \"https://localhost:8080/\", \"height\": 473} id=\"IvgzW71Yl0TH\" outputId=\"eef75ab9-a1e8-42e5-921e-58c80de2c505\"\ndf.dropna()\n\n# + id=\"GDRjVrRvmGW8\"\n\nfrom sklearn.cluster import KMeans\nimport datetime \nfrom dateutil.relativedelta import relativedelta\n\npd.set_option('display.max_columns', 100)\n\n# + id=\"32DPcHm3mSqb\"\nfrom sklearn.preprocessing import RobustScaler\nfrom sklearn.cluster import KMeans\nfrom sklearn import metrics\n\n\n# + [markdown] id=\"VNuQSSjjD9hV\"\n# **Calculate the performance of clustering using SSE and Sihouette score.**\n\n# + id=\"0b4gh4_cmZ-G\"\ndef plot_cluster(df, max_loop=50):\n \"\"\"\n Looking at the performance of various number of clusters using K-Means.\n Performance is evaluated by within cluster SSE and silhouette score.\n \"\"\"\n try:\n df.drop('cluster', axis=1, inplace=True)\n except:\n next\n X = df.iloc[:,[0,1]]\n \n # robust scaling is used so that the centering and scaling statistics are therefore not influenced by a few number of very large marginal outliers as they are based on percentiles\n rb = RobustScaler()\n X_rb = rb.fit_transform(X)\n \n sse_within_cluster = {}\n silhouette_score = {}\n \n for k in range(2, max_loop):\n kmeans = KMeans(n_clusters=k, random_state=10, n_init=10, n_jobs=-1)\n kmeans.fit(X_rb)\n sse_within_cluster[k] = kmeans.inertia_\n silhouette_score[k] = metrics.silhouette_score(X_rb, kmeans.labels_, random_state=10)\n\n _ = plt.figure(figsize=(10,6))\n ax1 = plt.subplot(211)\n _ = plt.plot(list(sse_within_cluster.keys()), list(sse_within_cluster.values()))\n _ = plt.xlabel(\"Number of Clusters\")\n _ = plt.ylabel(\"SSE Within Cluster\")\n _ = plt.title(\"Within Cluster SSE After K-Means Clustering\")\n _ = plt.xticks([i for i in range(2, max_loop)], rotation=75)\n \n ax2 = plt.subplot(212)\n _ = plt.plot(list(silhouette_score.keys()), list(silhouette_score.values()))\n _ = plt.xlabel(\"Number of Clusters\")\n _ = plt.ylabel(\"Silhouette Score\")\n _ = plt.title(\"Silhouette Score After K-Means Clustering\")\n _ = plt.xticks([i for i in range(2, max_loop)], rotation=75)\n \n plt.subplots_adjust(top=0.92, bottom=0.08, left=0.10, right=0.95, hspace=0.5, wspace=0.35)\n\n\n# + colab={\"base_uri\": \"https://localhost:8080/\", \"height\": 446} id=\"WVeezq2arGbu\" outputId=\"9a506994-fd8d-442d-9151-f553ac0e6237\"\nplot_cluster(df, max_loop=25)\n\n\n# + id=\"5frKfzztrvmy\"\ndef apply_cluster(df, clusters=6):\n \"\"\"\n Applying K-Means with the optimal number of clusters identified\n \"\"\"\n try:\n df.drop('cluster', axis=1, inplace=True)\n except:\n next\n X = df.iloc[:,[0,79]]\n rb = RobustScaler()\n X_rb = rb.fit_transform(X)\n kmeans = KMeans(n_clusters=clusters, random_state=17, n_init=10, n_jobs=-1) \n kmeans.fit(X_rb) \n score = metrics.silhouette_score(X_rb, kmeans.labels_, random_state=17)\n df['cluster'] = kmeans.labels_\n sse_within_cluster = kmeans.inertia_\n \n print(\"clustering performance\")\n print(\"sse withing cluster: \" + str(sse_within_cluster.round()))\n print(\"silhouette score: \" + str(score.round(2)))\n return df\n\n\n# + [markdown] id=\"NUdxPdSgGZOp\"\n# **First trial with 6 clusters**\n\n# + colab={\"base_uri\": \"https://localhost:8080/\"} id=\"Px0PU4Zer2IA\" outputId=\"3108ce94-f1ce-4eeb-efd9-995a77f12e5f\"\nfirst_trial = apply_cluster(df, clusters=6)\n\n# + colab={\"base_uri\": \"https://localhost:8080/\", \"height\": 235} id=\"KJHt14ZLsagd\" outputId=\"c0ed8fe9-326c-4367-da4d-86ada893cb60\"\ncluster_perf_df = (\n first_trial\n .groupby('cluster')\n .agg({\"tBodyAcc-mean()-X\":\"mean\", \"tBodyAcc-mean()-Y\":\"mean\"})\n .sort_values('tBodyAcc-mean()-X')\n .reset_index()\n)\n\ncluster_perf_df\n\n# + id=\"t6hJdvdFs-G-\"\ndf_sub = df.query(\"cluster == 0\").reset_index(drop=True)\n\n# + colab={\"base_uri\": \"https://localhost:8080/\", \"height\": 446} id=\"76cGm8_etD2l\" outputId=\"6dab758a-af41-4c8b-cfd1-4a0a88d0a136\"\nplot_cluster(df_sub, max_loop=25)\n\n# + [markdown] id=\"ypvXrZ3HHHLp\"\n# **Second Trial with 2 clusters**\n\n# + colab={\"base_uri\": \"https://localhost:8080/\"} id=\"6D3erPRMtW_6\" outputId=\"37ddfb07-81d1-4e1c-ef30-d2b23befc985\"\nsecond_trial= apply_cluster(df_sub, clusters=2)\n\n# + colab={\"base_uri\": \"https://localhost:8080/\", \"height\": 111} id=\"V9DgBmjmxG9f\" outputId=\"c228f115-188f-4186-c92e-064d77b19716\"\ncluster_perf_df = (\n second_trial\n .groupby('cluster')\n .agg({\"tBodyAcc-mean()-X\":\"mean\", \"tBodyAcc-mean()-Y\":\"mean\"})\n .sort_values('tBodyAcc-mean()-X')\n .reset_index()\n)\n\ncluster_perf_df\n\n# + [markdown] id=\"7u447gxzIkFS\"\n# The Silhouette score and SSE values for the trial first and second trial when compared it is founnd that the SSE score has dropped from 8731 to 397\n# where as Sihouette score has decreased a bit from 0.55 to 0.35.\n#\n# From the above we can conclude that cluster 2 is the golden cluster.\n","repo_name":"yadnyshree/ML_Assignments","sub_path":"Fractal_Clustering.ipynb","file_name":"Fractal_Clustering.ipynb","file_ext":"py","file_size_in_byte":5683,"program_lang":"python","lang":"en","doc_type":"code","stars":0,"dataset":"github-jupyter-script","pt":"42"} +{"seq_id":"32301327340","text":"# # intro to df 3 module + import dataset\n\nimport pandas as pd\n\nbond=pd.read_csv('jamesbond.csv')\nbond.head()\n\n# # set_index and reset_index methods\n\n# + active=\"\"\n# bond=pd.read_csv('jamesbond.csv',index_col='Film')\n# bond.head(2)\n# -\n\nbond.set_index(keys='Film',inplace=True)\nbond.head(3)\n\nbond.reset_index(inplace=True)\nbond.head(2)\n\n# # retrieve rows by index label with .loc[] accessor\n\nbond=pd.read_csv('jamesbond.csv',index_col='Film')\nbond.sort_index(inplace=True)\nbond.head(3)\n\nbond.loc['A View to a Kill']\n\nbond.loc['Casino Royale']\n\nbond.loc['Diamonds Are Forever':'From Russia With Love']\n\nbond.loc['Diamonds Are Forever':'From Russia With Love':2]\n\nbond.loc['Golden Eye': ]\n\nbond.loc[:'On Her Majesty a Secret Service']\n\nbond.loc[['Die Another Day','Octopussy']]\n\n# +\n#to check whether the index value is present in df\n\n'gold bond' in bond.index\n# -\n\n# # retrieve rows by index position with iloc accessor\n\nbond=pd.read_csv('jamesbond.csv')\nbond.head(3)\n\nbond.iloc[0]\n\nbond.iloc[[0,2]]\n\nbond.iloc[2:8]\nbond.iloc[0:4,2:4] #-----iloc[r,c]\n\n# + active=\"\"\n# loc= label indexing\n# iloc=integer indexing\n# -\n\n# # second arguments to loc and iloc accessors\n\nbond=pd.read_csv('jamesbond.csv',index_col='Film')\nbond.sort_index(inplace=True)\nbond.head(3)\n\nbond.loc['Moonraker','Actor']\nbond.loc['Moonraker','Director']\nbond.loc['Moonraker','Box Office']\n\nbond.loc['Moonraker',['Actor','Director','Box Office']]\n\nbond.loc[['Moonraker','A View to a Kill'],['Actor','Director','Box Office']]\n\nbond.loc['Moonraker','Actor':'Director']\n\nbond.loc['Moonraker':'Skyfall','Actor':'Box Office']\n\nbond.loc['Moonraker': ,'Director':]\n\nbond.iloc[12,3]\n\nbond.iloc[12,3:6]\n\nbond.iloc[[12,15],3:6]\n\nbond.iloc[2:,:5]\n\n# # set a new value for a specific cell\n\nbond=pd.read_csv('jamesbond.csv',index_col='Film')\nbond.sort_index(inplace=True)\nbond.head(3)\n\nbond.loc['Dr. No','Actor'] ='sir sean conery'\n\nbond.loc['Dr. No','Actor']\n\nbond.loc['Dr. No',['Box Office','Budget','Bond Actor Salary']] =[35,2,456]\n\nbond.loc['Dr. No',['Box Office','Budget','Bond Actor Salary']]\n\n# # set multiple value in a dataframe\n\nbond=pd.read_csv('jamesbond.csv',index_col='Film')\nbond.sort_index(inplace=True)\nbond.head(3)\n\nactor_is_seancornery=bond['Actor']=='Sean Connery'\nactor_is_seancornery\n\nbond.loc[actor_is_seancornery]\nbond.loc[actor_is_seancornery,'Actor']\nbond.loc[actor_is_seancornery,'Actor']= 'sir sean cornery'\n\nbond\n\n# # rename index labels or columns in a df\n\nbond=pd.read_csv('jamesbond.csv',index_col='Film')\nbond.sort_index(inplace=True)\nbond.head(3)\n\nbond.rename(mapper={'GoldenEye':'Golden Eye',\n 'A View to a Kill':'AViewToaKill'})\n\nbond.rename(mapper={'GoldenEye':'Golden Eye',\n 'A View to a Kill':'AViewToaKill'},axis=0)\nbond.rename(mapper={'GoldenEye':'Golden Eye',\n 'A View to a Kill':'AViewToaKill'},axis='rows')\nbond.rename(mapper={'GoldenEye':'Golden Eye',\n 'A View to a Kill':'AViewToaKill'},axis='index')\n\nbond.rename(index={'GoldenEye':'Golden Eye',\n 'A View to a Kill':'AViewToaKill'})\n\nbond.rename(mapper={'Year':'release date',\n 'Box Office':'revenue' },axis=1)\n#bond.rename(mapper={'Year':'release date',\n# 'Box Office':'revenue' },axis=columns)\n\nbond.columns=['year','hero','director','revenue','cost','salary']\n\nbond.head(0)\n\n# # delete rows or columns from a dataframe\n\nbond=pd.read_csv('jamesbond.csv',index_col='Film')\nbond.sort_index(inplace=True)\nbond.head(3)\n\nbond.drop('A View to a Kill')\nbond.drop(['Casino Royale','Diamonds Are Forever','Dr. No'])\n#bond.drop('A View to a Kill',inplace=True)\n\nbond.drop('Box Office',axis=1)\nbond.drop(['Box Office','Year'],axis=1)\n#bond.drop(['Box Office','Year'],axis=1,inplace=True)\n\nactor=bond.pop('Actor')\n\nactor\n\ndel bond['Director']\n\nbond\n\n# # create random sample\n\nbond=pd.read_csv('jamesbond.csv',index_col='Film')\nbond.sort_index(inplace=True)\nbond.head(3)\n\nbond.sample()\nbond.sample(n=4)\nbond.sample(frac=.25)\nbond.sample(n=3,axis=1)\n\n# # the nsmallest() and nlargest() methods\n\nbond=pd.read_csv('jamesbond.csv',index_col='Film')\nbond.sort_index(inplace=True)\nbond.head(3)\n\nbond.sort_values('Box Office',ascending=False).head(3)\n\n# +\nbond.nlargest(3,'Box Office')\n\nbond.nsmallest(3,'Box Office')\n# -\n\nbond['Budget'].nlargest(2)\n\n# # filtering with the where method\n\nbond=pd.read_csv('jamesbond.csv',index_col='Film')\nbond.sort_index(inplace=True)\nbond.head(3)\n\nmask=bond['Actor'] == 'Sean Connery'\nbond[mask]\n\nbond.where(mask)\n\nbond.where(bond['Box Office']>800)\n\nmask2=bond['Box Office']>800\n\n# +\n#bond.where(mask & mask2)\n# -\n\n# # query()method\n\nbond=pd.read_csv('jamesbond.csv',index_col='Film')\nbond.sort_index(inplace=True)\nbond.head(3)\n\nbond.columns=bond.columns.str.replace(' ','_')\n\nbond.query('Actor==\"Sean Connery\"')\nbond.query('Director==\"Terence Young\"')\nbond.query('Director!=\"Terence Young\"')\n\nbond.query('Box_Office > 500')\n\nbond.query(\"Actor =='Roger Moore' or Director =='Terence Young' \")\n\n# +\nbond.query(\"Actor in ['Roger Moore','Sean Connery']\")\n\n#bond.query(\"Actor not in ['Roger Moore','Sean Connery']\")\n\n# + active=\"\"\n# to run query method there should be no space between the column names\n# -\n\n# # review of apply method on a single column\n\nbond=pd.read_csv('jamesbond.csv',index_col='Film')\nbond.sort_index(inplace=True)\nbond.head(3)\n\n\n# +\ndef convert_to_string_and_millions(number):\n return str(number) +'millions'\n\n#bond['Box Office']=bond['Box Office'].apply(convert_to_string_and_millions)\n\n\n# +\n#bond['Budget']=bond['Budget'].apply(convert_to_string_and_millions)\n# -\n\nc=['Box Office','Budget','Bond Actor Salary']\nfor i in c:\n bond[i]=bond[i].apply(convert_to_string_and_millions)\n\nbond.head(2)\n\n# # apply method with row values\n\nbond=pd.read_csv('jamesbond.csv',index_col='Film')\nbond.sort_index(inplace=True)\nbond.head(3)\n\n\n# +\ndef good_movie(row):\n actor= row[1]\n budget=row[4]\n \n if actor == 'Roger Moore':\n return 'best'\n elif actor == 'Daniel Craig' and budget >40:\n return 'ok'\n else:\n return 'no clue'\n \nbond.apply(good_movie,axis=1)\n# -\n\n# # # copy method\n\nbond=pd.read_csv('jamesbond.csv',index_col='Film')\nbond.sort_index(inplace=True)\nbond.head(3)\n\ndirectors=bond['Director']\ndirectors.head(3)\n\n# +\n#directors['A View to a Kill']='mister john glen'\n\n# +\n#directors.head(3)\n\n# +\n#bond.head(3)\n# -\n\ndirectors=bond['Director'].copy()\ndirectors.head(3)\n\ndirectors['A View to a Kill']='mister john glen'\n\ndirectors.head(3)\n\nbond.head(3)\n","repo_name":"subhaganesh/data-analysis_using_pandas","sub_path":"dataframes 3 - extraction.ipynb","file_name":"dataframes 3 - extraction.ipynb","file_ext":"py","file_size_in_byte":6512,"program_lang":"python","lang":"en","doc_type":"code","stars":0,"dataset":"github-jupyter-script","pt":"42"} +{"seq_id":"34676164251","text":"# +\n#Q1 program to find a number divisible by 7 and not a multiple of 5 ranging between 2000 to 3200\n\nnumber = []\nfor i in range(2000, 3201):\n if (i % 7 == 0) and (i % 5 != 0):\n number.append(i)\nprint(number)\n\n# +\n#Q2 accept users's first and last name and print in a reverse order\n\nfname = input(\"enter your first name: \")\nlname = input(\"enter your last name: \")\nprint(lname + \" \" + fname)\n\n# +\n#Q3 find the volume of a sphere where diameter = 12 cm\n\ndiameter = 12\nradius = diameter/2\nvolume = 4/3 * 3.14 * radius**3\nprint(\"The volume of the sphere is\", volume)\n# -\n\n\n","repo_name":"AvirupM99/PythonAssignments","sub_path":"PythonAssignment1.ipynb","file_name":"PythonAssignment1.ipynb","file_ext":"py","file_size_in_byte":579,"program_lang":"python","lang":"en","doc_type":"code","stars":0,"dataset":"github-jupyter-script","pt":"42"} +{"seq_id":"29621788749","text":"## Problem1\nimport random as rd\nimport numpy as np\nimport matplotlib.pyplot as plt\n\n#input parameters\nprice = 75\ncost = 50\nsalvage = 15\ndemand_mean = 500\ndemand_std = 75\norder = 550\n\n#base case model\ndemand = 450 #we assume this number for demand\nif demand < order: qty_sold = demand\nelse: qty_sold = order\nqty_left = order - qty_sold\nsales_revenue = price * qty_sold\nsalvage_revenue = salvage * qty_left\ntotal_cost = cost * order\nprofit = sales_revenue + salvage_revenue - total_cost\nprint(qty_sold, qty_left, sales_revenue, salvage_revenue, total_cost, profit)\n\n# +\n#simulation trials\nrd.seed(101010)\ntrials = 10000\nsample = list()\nfor i in range(trials):\n demand = rd.normalvariate(demand_mean, demand_std)\n if demand < order: qty_sold = demand \n else: qty_sold = order\n qty_left = order - qty_sold\n sales_revenue = price * qty_sold\n salavage_revenue = salvage * qty_left\n total_cost = cost * order\n profit = sales_revenue + salvage_revenue - total_cost\n sample.append(profit)\n \n(a,b,c) = plt.hist(sample, edgecolor='k')\n# -\n\nmean = sum(sample)/len(sample)\nprint('Mean = $%5.2f' % (sum(sample)/len(sample)))\n\nproportion = sum(1 for x in sample if x < 0.0)/len(sample)\nprint('There is a %5.2f%% chance of incurring a loss on the order.'% (100*proportion))\n\nprint(sample[0:10])\n\n# part(d)\norder_1st = list(range(300,801))\nmean_profit_1st = list()\nfor order in order_1st:\n trials = 10000\n sample = list()\n for i in range(trials):\n demand = rd.normalvariate(demand_mean, demand_std)\n if demand < order: qty_sold = demand \n else: qty_sold = order\n qty_left = order - qty_sold\n sales_revenue = price * qty_sold\n salavage_revenue = salvage * qty_left\n total_cost = cost * order\n profit = sales_revenue + salvage_revenue - total_cost\n sample.append(profit)\n mean_profit = sum(sample)/len(sample)\n mean_profit_1st.append(mean_profit)\n\n\nplt.plot(order_1st, mean_profit_1st)\n\nprint('The optimal order quantity is %d.' % (order_1st[mean_profit_1st.index(max(mean_profit_1st))]))\nprint('The expected profit is $%8.2f.'% (max(mean_profit_1st)))\n\n# +\n##Question 2\n\n#input parameters\nprice_min = 18.95\nprice_max = 26.95\nprice_mode = 24.95\ncost_min = 12.00\ncost_max = 15.00\nintercept = 10000.00\nslope = -250.00\nrand_term_mean = 0.00\nrand_term_std = 10.00\nfixed_cost_mean = 30000\nfixed_cost_std = 5000\n# -\n\n#base case model\nprice = 24.95\ncost = 13.50\nrand_term = -10.00\nquantity_sold = intercept + slope*price + rand_term\nfixed_cost = 30000\nprofit = (price-cost)*quantity_sold - fixed_cost\nprint(quantity_sold, profit)\n\n#simulation\ntrials = 10000\nsample = list()\nfor i in range(trials):\n price = rd.triangular(price_min, price_max, price_mode)\n cost = rd.uniform(cost_min, cost_max)\n rand_term = rd.normalvariate(rand_term_mean, rand_term_std)\n quantity_sold = intercept + slope*price + rand_term\n fixed_cost = rd.normalvariate(fixed_cost_mean, fixed_cost_std)\n profit = (price - cost) * quantity_sold - fixed_cost\n sample.append(profit)\nplt.hist(sample, edgecolor =\"k\")\n\nprint('The expected mean profit is $%8.2f' % (sum(sample)/len(sample)))\nproportion = sum(1 for x in sample if x < 0.0)/len(sample)\nprint('The probability of incurring a lost is %5.2f%%' % (100*proportion))\nprint('The maximum lost is $%5.2f.' % (min(sample)))\n\n# +\n#2a Profit appears to be normally distributed\n","repo_name":"leah1217/Decision-Models","sub_path":"Homework 4 - Leah Ngan Lai.ipynb","file_name":"Homework 4 - Leah Ngan Lai.ipynb","file_ext":"py","file_size_in_byte":3396,"program_lang":"python","lang":"en","doc_type":"code","stars":0,"dataset":"github-jupyter-script","pt":"42"} +{"seq_id":"28421349266","text":"# +\nlist_of_numbers = [3, 5, 6, 8, 10, 11, 25, 55, 95]\n\n# функция, для нахождения минимума в списке целых\ndef min_value(some_list):\n number = min(some_list)\n return number\nprint(f'The minimum value in the list is {min_value(list_of_numbers)}')\n\n\n# найти количество простых чисел в списке целых\n\ndef get_prime_number(some_list):\n count = 0\n for i in some_list:\n if all(i % j != 0 for j in range(2, i)):\n count += 1\n return count\n \nprint(f'The quantity of prime numbers = {get_prime_number(list_of_numbers)}')\n\n\n# функция, удаляющую из списка целых некоторое заданное число\n\ndef get_delete_number(some_list, number):\n count = 0\n for i in some_list:\n if i == number:\n some_list.remove(i)\n count += 1\n return count\n\nget_delete_number(list_of_numbers, 10)\nprint(list_of_numbers) \n\n# функция, которая возвращает список, содержащий элементы 2-х списков\n\nlist_of_numbers_1 = [3, 5, 6, 8, 10, 25, 55, 95]\nlist_of_numbers_2 = [56, 89, 23, 54]\n\ndef get_full_list(some_list, some_list_2):\n return some_list + some_list_2\n\nprint(get_full_list(list_of_numbers_1, list_of_numbers_2))\n\n# функция, высчитывающуя степень каждого элемента списка целых\n\ndef get_power_of_numbers(some_list, pow_number):\n result_list = []\n for i in some_list:\n result_list.append(i**pow_number)\n return result_list\n\nprint(get_power_of_numbers(list_of_numbers, 2))\n \n \n","repo_name":"nadiia-tokareva/skillup","sub_path":"lesson 3-1.ipynb","file_name":"lesson 3-1.ipynb","file_ext":"py","file_size_in_byte":1694,"program_lang":"python","lang":"en","doc_type":"code","stars":0,"dataset":"github-jupyter-script","pt":"42"} +{"seq_id":"33475372493","text":"import numpy as np\n\nimport pandas as pd\nsales = pd.read_csv('./FreeCodeCamp-Pandas-Real-Life-Example-master/data/sales_data.csv',parse_dates=['Date'])\nsales.head()\n\nsales.describe()\n\nsales.shape\n\nsales.info()\n\nsales['Unit_Cost'].mean()\n\n# +\n#plot a box plot using the selected feature\n\nimport matplotlib.pyplot as plt \nplt.rcParams[\"font.family\"] = \"sans serif\"\n\nsales['Unit_Cost'].plot(kind='box', vert= False, figsize=(14,6))\n\n \n\n# +\nimport seaborn as sns\n\nplt.figure(figsize=(10,6))\nsns.kdeplot(data=sales, x=sales.Unit_Cost)\nplt.xlabel('Unit Cost')\nplt.show()\n# -\n\n\n\nsales['Age_Group'].value_counts()\n\nsales['Age_Group'].value_counts().plot(kind='pie', figsize=(6,6))\n\nax = sales['Age_Group'].value_counts().plot(kind='bar', figsize=(14,6))\nax.set_ylabel('Number of Sales')\nax.set_xlabel('Age Group')\n\nnew_sales = sales.select_dtypes(include=np.number)\ncorr = new_sales.corr()\nsns.heatmap(corr, annot=False)\nplt.show()\n\ncorr\n\n# +\nimport plotly.express as px\n\npx.line(x=[1,2,3,4], y=[4,5,6,7]).show()\n# -\n\n# !pip install nbformat\n","repo_name":"SebasManco/Projects","sub_path":"MachineLearning/DataScience/.ipynb_checkpoints/my_lecture1-checkpoint.ipynb","file_name":"my_lecture1-checkpoint.ipynb","file_ext":"py","file_size_in_byte":1035,"program_lang":"python","lang":"en","doc_type":"code","stars":0,"dataset":"github-jupyter-script","pt":"42"} +{"seq_id":"72060773566","text":"import pandas as pd\nfrom sklearn.metrics import mean_squared_error\nfrom math import sqrt\nfrom sklearn.ensemble import RandomForestRegressor\nimport numpy as np\nfrom sklearn.metrics import r2_score\nfrom sklearn.model_selection import train_test_split\nfrom sklearn.datasets import load_breast_cancer\nimport sys\nfrom sklearn import tree\nimport matplotlib.pyplot as plt\nimport graphviz\n\n\n# +\ndf = pd.read_csv('train.csv')\ndf_train, df_test = train_test_split(df,test_size = 0.2,shuffle=False)\nfeature_names = ['OverallQual', 'GrLivArea', 'GarageCars']\nX_train = df_train[feature_names]\ny_train = df_train['SalePrice']\ny_train.to_csv('y_train.csv')\n\nX_test = df_test[feature_names]\ny_test = df_test[['SalePrice']]\n# -\n\nX, y = load_breast_cancer(return_X_y=True)\nX_train, X_test, y_train, y_test = train_test_split(X, y, random_state=0)\n\n# +\n## build sklearn regression trree\ntree_max_depth = 5 # controls overfitting \nmin_samples_leaf = 5 # controls overfitting \n\nsk_dt_reg = tree.DecisionTreeRegressor(max_depth = tree_max_depth, min_samples_leaf = min_samples_leaf, criterion = 'squared_error')\nsk_dt_reg.fit(X_train,y_train)\nsk_dt_preds = sk_dt_reg.predict(X_train)\nsk_dt_scores = r2_score(sk_dt_preds,y_train)\nprint('sk dt scores train:',sk_dt_scores)\nsk_dt_preds = sk_dt_reg.predict(X_test)\nsk_dt_scores = r2_score(sk_dt_preds,y_test)\nprint('sk dt scores test:',sk_dt_scores)\n# -\n\n### plot the tree\ndot_data = tree.export_graphviz(sk_dt_reg,out_file = None,feature_names = feature_names, filled = True, special_characters = True,max_depth=3)\ngraph = graphviz.Source(dot_data)\n# graph\n\n# +\n## build sklearn random forest\nn_estimators = 100\ncriterion='squared_error'\nmax_depth = None\nmin_samples_split = 2\nmin_samples_leaf = 1\nmin_weight_fraction_leaf = 0\nmax_features = sqrt(len(feature_names) - 1)/len(feature_names) # according to GENIE3 suggestion\nmax_leaf_nodes = None\nmin_impurity_decrease = 0\nbootstrap = True\noob_score = True # to use out-of-bag samples to estimate the generalization score (available for bootstrap=True)\nn_jobs = None # number of jobs in parallel. fit, predict, decision_path, and apply can be done in parallel over the trees\nrandom_state=None # controls randomness in bootstrapping as well as drawing features\nverbose=1 \nwarm_start=False # reuse the slution of the previous call to fit and add more ensembles to the estimator. look up on Glosery\nccp_alpha=0 # complexity parameter used for minima cost-complexity pruning. by default, no prunning\nmax_samples = None # if bootstrap is True, the number of samples to draw from the samples. if none, draw X.shape[0]\n\nsk_dt_reg = RandomForestRegressor(n_estimators=n_estimators, criterion=criterion, max_depth=max_depth, min_samples_split=min_samples_split,\n min_samples_leaf=min_samples_leaf, min_weight_fraction_leaf=min_weight_fraction_leaf, \n max_features=max_features, max_leaf_nodes=max_leaf_nodes, min_impurity_decrease=min_impurity_decrease,\n bootstrap=bootstrap, oob_score=oob_score, n_jobs=n_jobs, random_state=random_state, verbose=verbose,\n warm_start=warm_start, ccp_alpha=ccp_alpha, max_samples=max_samples)\nsk_dt_reg.fit(X_train,y_train)\nsk_dt_preds = sk_dt_reg.predict(X_train)\nsk_dt_scores = r2_score(sk_dt_preds,y_train)\nprint('sk dt scores train:',sk_dt_scores)\nsk_dt_preds = sk_dt_reg.predict(X_test)\nsk_dt_scores = r2_score(sk_dt_preds,y_test)\nprint('sk dt scores test:',sk_dt_scores)\n# -\n\n### build sklearn random forest\nsk_dt_reg = RandomForestRegressor()\nsk_dt_reg.fit(X_train,y_train)\nsk_dt_preds = sk_dt_reg.predict(X_train)\nsk_dt_scores = r2_score(sk_dt_preds,y_train)\nprint('sk dt scores train:',sk_dt_scores)\nsk_dt_preds = sk_dt_reg.predict(X_test)\nsk_dt_scores = r2_score(sk_dt_preds,y_test)\nprint('sk dt scores test:',sk_dt_scores)\n\ndot_data = tree.export_graphviz(clf, out_file=None, \n feature_names=iris.feature_names, \n class_names=iris.target_names, \n filled=True, rounded=True, \n special_characters=True) \ngraph = graphviz.Source(dot_data) \n\n\nclass Node:\n def __init__(self, x, y, idxs, min_leaf=5):\n self.x = x \n self.y = y\n self.idxs = idxs \n self.min_leaf = min_leaf\n self.row_count = len(idxs)\n self.col_count = x.shape[1]\n self.val = np.mean(y[idxs]) \n self.score = float('inf')\n self.find_varsplit()\n def find_varsplit(self): #find where to split\n for c in range(self.col_count): self.find_better_split(c) # after this, the row and column of split is determined by scoring\n if self.is_leaf: return\n x = self.split_col\n lhs = np.nonzero(x <= self.split)[0]\n rhs = np.nonzero(x > self.split)[0]\n self.lhs = Node(self.x, self.y, self.idxs[lhs], self.min_leaf)\n self.rhs = Node(self.x, self.y, self.idxs[rhs], self.min_leaf)\n @property\n def split_col(self): return self.x.values[self.idxs,self.var_idx]\n\n @property\n def is_leaf(self): return self.score == float('inf') \n def find_better_split(self, var_idx): # determines row and column of the split\n x = self.x.values[self.idxs, var_idx]\n for r in range(self.row_count):\n lhs = x <= x[r]\n rhs = x > x[r]\n if rhs.sum() < self.min_leaf or lhs.sum() < self.min_leaf: continue # prunning\n\n curr_score = self.find_score(lhs, rhs)\n if curr_score < self.score: \n self.var_idx = var_idx # the chosen column\n self.score = curr_score\n self.split = x[r] # the row to split\n\n def find_score(self, lhs, rhs):\n y = self.y[self.idxs]\n lhs_std = y[lhs].std()\n rhs_std = y[rhs].std()\n# return lhs_std * lhs.sum() + rhs_std * rhs.sum() #score is calculated by: std_l * sum_x_l + std_r*sum_x_r => ??? why sum of ihs? \n return lhs_std + rhs_std #score is calculated by: std_l * sum_x_l + std_r*sum_x_r => ??? why sum of ihs? \n\n def predict(self, x):\n return np.array([self.predict_row(xi) for xi in x])\n\n def predict_row(self, xi):\n if self.is_leaf: return self.val\n node = self.lhs if xi[self.var_idx] <= self.split else self.rhs\n return node.predict_row(xi)\nclass DecisionTreeRegressor:\n \n def fit(self, X, y, min_leaf = 5):\n self.dtree = Node(X, y, idxs = np.array(np.arange(len(y))), min_leaf=min_leaf)\n return self\n\n def predict(self, X):\n return self.dtree.predict(X.values)\nregressor = DecisionTreeRegressor().fit(X_train, y_train)\n\npreds = regressor.predict(X)\nr2_score(y, preds)\n\npred_test = regressor.predict(X_test)\nr2_score(y_test, pred_test)\n","repo_name":"janursa/DT","sub_path":"script.ipynb","file_name":"script.ipynb","file_ext":"py","file_size_in_byte":6763,"program_lang":"python","lang":"en","doc_type":"code","stars":0,"dataset":"github-jupyter-script","pt":"42"} +{"seq_id":"19206124922","text":"# ### Author - Nikita Koshti\n\n# # GRIP - THE SPARKS FOUNDATION\n\n# ### DATA SCIENCE AND BUSINESS ANALYTICS INTERNSHIP\n\n# (BATCH-APRIL 2021)\n\n# ### Prediction using unsupervised learning\n\n# ### Task-2 predict the optimum number of clusters and represent it visually from the given iris data.\n\n# In this model we will find number of cluster in iris data by k-Means clustering algorithim\n\n# ### importing and reading data\n\n#importing liabraries\nimport pandas as pd\nimport numpy as np\nimport matplotlib.pyplot as plt\nimport seaborn as sns\nimport warnings\nwarnings.filterwarnings(\"ignore\")\n\n#read the data\ndata=pd.read_csv(\"Iris.csv\")\ndata.head()\n\n#data information\ndata.shape\n\n#data information\ndata.info()\n\n#Describing data\ndata.describe()\n\ndata['Species'].value_counts()\n\n#checking null value\ndata.isnull().sum()\n\n# + active=\"\"\n# We can see there are no null value in dataset\n# -\n\n#Droping ID column and creating plot\ntmp=data.drop('Id', axis=1)\ng=sns.pairplot(tmp,hue='Species',markers='*')\nplt.show()\n\n# ### Findout cluster with KMeans clustering\n\n# +\nx=data.iloc[:,[0,1,2,3]].values\nfrom sklearn.cluster import KMeans\nwcss = [] #within cluster sum of square\n\nfor i in range(1,10):\n kmeans = KMeans(n_clusters = i, init ='k-means++',\n max_iter = 300, n_init = 10, random_state = 0)\n kmeans.fit(x)\n wcss.append(kmeans.inertia_)\nwcss \n# -\n\n#ploting Graph\nplt.figure(figsize=(9,6))\nplt.plot(range(1,10),wcss)\nplt.title('The Elbow method')\nplt.xlabel('Number of cluster')\nplt.ylabel('cluster')\nplt.show()\n\n# From the above graph we can find out the optimum number of clusters to be 3 as the elbow starts having a straight trend after 3\n\n#Appling Kmeans\nkmeans= KMeans(n_clusters = 3, init = 'k-means++',\n max_iter = 300, n_init = 10, random_state = 0)\ny=kmeans.fit_predict(x)\ny\n\ncl=pd.Series(kmeans.labels_)\ndata['cluster']=cl\n\n# ## visualising the clusters\n\n# +\nplt.figure(figsize=(9,6))\nplt.scatter(x[y == 0,0],x[y == 0,1],\n s= 100, c = 'green', label = 'Iris-setosa')\nplt.scatter(x[y == 1,0],x[y == 1,1],\n s= 100, c = 'red', label = 'Iris-versicolour')\nplt.scatter(x[y == 2,0],x[y == 2,1],\n s= 100, c = 'blue', label = 'Iris-virginica')\n\n#ploting the centroids of the clusters\nplt.scatter(kmeans.cluster_centers_[:,0],kmeans.cluster_centers_[:,1],\n s= 100, c = 'black', label = 'Centroids')\nplt.legend()\nplt.show()\n\n# +\n#visualising the clusters - On the first two columns\nplt.figure(figsize=(9,6))\nplt.scatter(x[y == 0,0],x[y == 0,2],\n s= 100, c = 'green', label = 'Iris-setosa')\nplt.scatter(x[y == 1,1],x[y == 1,2],\n s= 100, c = 'red', label = 'Iris-versicolour')\nplt.scatter(x[y == 2,1],x[y == 2,2],\n s= 100, c = 'blue', label = 'Iris-virginica')\n\n#ploting the centroids of the clusters\nplt.scatter(kmeans.cluster_centers_[:,0],kmeans.cluster_centers_[:,1],\n s= 100, c = 'black', label = 'Centroids')\nplt.legend()\nplt.show()\n# -\n\n# ## Representing the data with the cluster column\n\ndata.iloc[:,:]\n\nx=[5.1,4.6,3.0,6.7]\nkmeans.predict([x])\n\n\n# Through we can assume that our model is predict accurately.\n","repo_name":"NIkita7898-nmk/Predict-the-optimum-number-of-cluster","sub_path":"Task-2 predict the optimum number of clusters and represent it visually.ipynb","file_name":"Task-2 predict the optimum number of clusters and represent it visually.ipynb","file_ext":"py","file_size_in_byte":3153,"program_lang":"python","lang":"en","doc_type":"code","stars":0,"dataset":"github-jupyter-script","pt":"42"} +{"seq_id":"39929629253","text":"# + [markdown] id=\"view-in-github\" colab_type=\"text\"\n#\n\n# + colab={\"base_uri\": \"https://localhost:8080/\"} id=\"GIvkYm2-7gCl\" outputId=\"f445fa96-9491-4b53-ecdc-cd378fb57659\"\n# Input total score for each subject\nmaths_score = int(input(\"Enter your overall marks for Maths: \"))\nphysics_score = int(input(\"Enter your overall marks for Physics: \"))\nchemistry_score = int(input(\"Enter your overall marks for Chemistry: \"))\n\n\n# + id=\"mIcD3Xy19jQl\"\n# Calculate the percentage for each subject\ntotal_marks = 100 # Assume the maximum marks for each subject is 100\nmaths_percentage = (maths_score / total_marks) * 100\nphysics_percentage = (physics_score / total_marks) * 100\nchemistry_percentage = (chemistry_score / total_marks) * 100\n\n# + colab={\"base_uri\": \"https://localhost:8080/\"} id=\"lgRkYiwD9puy\" outputId=\"5e86b15a-20e0-4db6-f8b9-3e6eb8c3ebe0\"\n# Print the results\nprint(f\"Maths - Total Score: {maths_score}, Percentage: {maths_percentage}%\")\nprint(f\"Physics - Total Score: {physics_score}, Percentage: {physics_percentage}%\")\nprint(f\"Chemistry - Total Score: {chemistry_score}, Percentage: {chemistry_percentage}%\")\n\n","repo_name":"atonendeley/IMAGE-STEGANOGRAPHY","sub_path":"Marks2.ipynb","file_name":"Marks2.ipynb","file_ext":"py","file_size_in_byte":1332,"program_lang":"python","lang":"en","doc_type":"code","stars":0,"dataset":"github-jupyter-script","pt":"42"}
+{"seq_id":"31758996470","text":"# # Week 3 Assignment - Clustering Neighborhoods in Toronto\n\n# +\nimport numpy as np # library to handle data in a vectorized manner\n\nimport pandas as pd # library for data analsysis\npd.set_option('display.max_columns', None)\npd.set_option('display.max_rows', None)\n\nimport json # library to handle JSON files\n\n# #!conda install -c conda-forge geopy --yes # uncomment this line if you haven't completed the Foursquare API lab\nfrom geopy.geocoders import Nominatim # convert an address into latitude and longitude values\n\nimport requests # library to handle requests\nimport bs4\nfrom bs4 import BeautifulSoup\nimport urllib.request\n\nfrom pandas.io.json import json_normalize # tranform JSON file into a pandas dataframe\n\n# Matplotlib and associated plotting modules\nimport matplotlib.cm as cm\nimport matplotlib.colors as colors\n\n# import k-means from clustering stage\nfrom sklearn.cluster import KMeans\n\n# #!conda install -c conda-forge folium=0.5.0 --yes # uncomment this line if you haven't completed the Foursquare API lab\nimport folium # map rendering library\n\nprint('Libraries imported.')\n# -\n\n# ## Pulling the Data from Wikipedia \n\ntables = pd.read_html('List_of_postal_codes_of_Canada:_M.html', header=0)\ndf=tables[0]\ndf = df[df.Borough != 'Not assigned']\n\ndf1=df.groupby('Postcode', as_index=False).agg({'Borough': 'first', 'Neighbourhood' : ' , '.join})\ndf1 = df1.sort_values(by='Postcode')\n\n# correct the not-assigned neighbourhood\ndf1['Neighbourhood'] = np.where(df1['Neighbourhood']=='Not assigned', df1['Borough'], df1['Neighbourhood'])\ndf1\n\ndf1.shape\n\n# ### As an American I cannot tell you how much time I wasted before I realized that \"Neighbourhood\" included a u. We spell it \"Neighborhood\".\n\n# # Part 2: Creating the file with the Lat/Long columns\n\n# ! pip install geocoder\n\nimport geocoder\ng = geocoder.google('Toronto, Ontario')\ng\n\n# ### I could not get even a single report from the geocoder, and wanted to get going with it, so I choose to use the csv.\n\nGS = pd.read_csv('Geospatial_Coordinates.csv')\ndf1.rename(columns={'Postcode' : 'Postal Code', 'Latitude':'Latitude'}, inplace=True)\n\ndf1=df1.merge(GS, on=\"Postal Code\")\ndf1.head()\n\n# # Part 3: Clustering Neighborhoods\n\n# +\naddress = 'Toronto, Ontario'\n\ngeolocator = Nominatim(user_agent=\"ny_explorer\")\nlocation = geolocator.geocode(address)\nlatitude = location.latitude\nlongitude = location.longitude\nprint('The geograpical coordinate of Toronto are {}, {}.'.format(latitude, longitude))\n\n# +\n# First let's make a map.\n# create map of Toronto using latitude and longitude values\nmap_toronto = folium.Map(location=[latitude, longitude], zoom_start=10)\nfor lat, lng, borough, neighborhood in zip(df1['Latitude'], df1['Longitude'], df1['Borough'], df1['Neighbourhood']):\n label = '{}, {}'.format(neighborhood, borough)\n label = folium.Popup(label, parse_html=True)\n folium.CircleMarker(\n [lat, lng],\n radius=5,\n popup=label,\n color='blue',\n fill=True,\n fill_color='#3186cc',\n fill_opacity=0.7,\n parse_html=False).add_to(map_toronto) \n \nmap_toronto\n# -\n\n# ## Get FourSquare Data\n\nCLIENT_ID = 'MJ2INACVWX1BIQZF4PZKQTRJKVYSBM0JHLEPZWPMOAMQPURV' # your Foursquare ID\nCLIENT_SECRET = '5TH5H3G5TOXAKNNGKR22LC2EWJAV5MD0UQBJHVQAZ1WHVEW2' # your Foursquare Secret\nVERSION = '20180605' # Foursquare API version\nLIMIT=100\n\n\ndef getNearbyVenues(names, latitudes, longitudes, radius=500):\n \n venues_list=[]\n for name, lat, lng in zip(names, latitudes, longitudes):\n print(name)\n \n # create the API request URL\n url = 'https://api.foursquare.com/v2/venues/explore?&client_id={}&client_secret={}&v={}&ll={},{}&radius={}&limit={}'.format(\n CLIENT_ID, \n CLIENT_SECRET, \n VERSION, \n lat, \n lng, \n radius, \n LIMIT)\n \n # make the GET request\n results = requests.get(url).json()[\"response\"]['groups'][0]['items']\n \n # return only relevant information for each nearby venue\n venues_list.append([(\n name, \n lat, \n lng, \n v['venue']['name'], \n v['venue']['location']['lat'], \n v['venue']['location']['lng'], \n v['venue']['categories'][0]['name']) for v in results])\n\n nearby_venues = pd.DataFrame([item for venue_list in venues_list for item in venue_list])\n nearby_venues.columns = ['Neighborhood', \n 'Neighborhood Latitude', \n 'Neighborhood Longitude', \n 'Venue', \n 'Venue Latitude', \n 'Venue Longitude', \n 'Venue Category']\n \n return(nearby_venues)\n\n\nToronto_venues = getNearbyVenues(names=df1['Neighbourhood'],\n latitudes=df1['Latitude'],\n longitudes=df1['Longitude']\n )\n\n\n# ### Obviously, this is running by postal code, not by \"neighbourhood\" as in the New York example.\n\nprint(Toronto_venues.shape)\nToronto_venues.head()\n\n# ## Look I'll be frank this isn't the kind of data I'm interested in using in my career, and I'm tired of mucking with GSP data, so here's the minimum to get to the clustering.\n\n# +\n# one hot encoding\ntoronto_onehot = pd.get_dummies(Toronto_venues[['Venue Category']], prefix=\"\", prefix_sep=\"\")\n\n# add neighborhood column back to dataframe\ntoronto_onehot['Neighborhood'] = Toronto_venues['Neighborhood'] \n\n# move neighborhood column to the first column\nfixed_columns = [toronto_onehot.columns[-1]] + list(toronto_onehot.columns[:-1])\ntoronto_onehot = toronto_onehot[fixed_columns]\n\ntoronto_onehot.head()\n# -\n\ntoronto_onehot.shape\n\ntoronto_grouped = toronto_onehot.groupby('Neighborhood').mean().reset_index()\ntoronto_grouped\n\ntoronto_grouped.shape\n\n\n# ## Toronto is a nice city, though I prefer Ottawa. I don't actually care about which venues are most common. Here's an implementation of sorting the data so that we are already prioritizing which venues will be viewed on the clustering algorithm. \n\ndef return_most_common_venues(row, num_top_venues):\n row_categories = row.iloc[1:]\n row_categories_sorted = row_categories.sort_values(ascending=False)\n \n return row_categories_sorted.index.values[0:num_top_venues]\n\n\n# +\nnum_top_venues = 10\n\nindicators = ['st', 'nd', 'rd']\n\n# create columns according to number of top venues\ncolumns = ['Neighborhood']\nfor ind in np.arange(num_top_venues):\n try:\n columns.append('{}{} Most Common Venue'.format(ind+1, indicators[ind]))\n except:\n columns.append('{}th Most Common Venue'.format(ind+1))\n\n# create a new dataframe\nneighborhoods_venues_sorted = pd.DataFrame(columns=columns)\nneighborhoods_venues_sorted['Neighborhood'] = toronto_grouped['Neighborhood']\n\nfor ind in np.arange(toronto_grouped.shape[0]):\n neighborhoods_venues_sorted.iloc[ind, 1:] = return_most_common_venues(toronto_grouped.iloc[ind, :], num_top_venues)\n\nneighborhoods_venues_sorted.shape\n# -\n\n# # Part Too Many: The actual clustering\n\n# +\n# set number of clusters\nkclusters = 5\n\ntoronto_grouped_clustering = toronto_grouped.drop('Neighborhood', 1)\n\n# run k-means clustering\nkmeans = KMeans(n_clusters=kclusters, random_state=0).fit(toronto_grouped_clustering)\n\n# check cluster labels generated for each row in the dataframe\nkmeans.labels_\n\n# +\n# add clustering labels\nneighborhoods_venues_sorted.insert(0, 'Cluster Labels', kmeans.labels_)\n\n\n# -\n\nneighborhoods_venues_sorted['Cluster Labels']\n\ndf1.rename(columns={'Neighbourhood' : 'Neighborhood'}, inplace=True)\n\n# +\ntoronto_merged = df1\n\n# merge toronto_grouped with toronto_data to add latitude/longitude for each neighborhood\ntoronto_merged = toronto_merged.join(neighborhoods_venues_sorted.set_index('Neighborhood'), on='Neighborhood')\ntoronto_merged.head()\n# -\n\ntoronto_merged.drop(toronto_merged.index[[16,20,21,93,98]], inplace=True)\ntoronto_merged['Cluster Labels']=toronto_merged['Cluster Labels'].astype(int)\ntoronto_merged.head()\n\n# set color scheme for the clusters\nx = np.arange(kclusters)\nys = [i + x + (i*x)**2 for i in range(kclusters)]\ncolors_array = cm.rainbow(np.linspace(0, 1, len(ys)))\nrainbow = [colors.rgb2hex(i) for i in colors_array]\n\n\n# +\n# create map\nmap_clusters = folium.Map(location=[latitude, longitude], zoom_start=11)\n\n# add markers to the map\nmarkers_colors = []\nfor lat, lon, poi, cluster in zip(toronto_merged['Latitude'], toronto_merged['Longitude'], toronto_merged['Neighborhood'], toronto_merged['Cluster Labels']):\n label = folium.Popup(str(poi) + ' Cluster ' + str(cluster), parse_html=True)\n folium.CircleMarker(\n [lat, lon],\n radius=5,\n popup=label,\n color=rainbow[cluster-1],\n fill=True,\n fill_color=rainbow[cluster-1],\n fill_opacity=0.7).add_to(map_clusters)\n \nmap_clusters\n# -\n\ntoronto_merged.loc[toronto_merged['Cluster Labels'] == 0, toronto_merged.columns[[1] + list(range(5, toronto_merged.shape[1]))]]\n\ntoronto_merged.loc[toronto_merged['Cluster Labels'] == 1, toronto_merged.columns[[1] + list(range(5, toronto_merged.shape[1]))]]\n\ntoronto_merged.loc[toronto_merged['Cluster Labels'] == 2, toronto_merged.columns[[1] + list(range(5, toronto_merged.shape[1]))]]\n\ntoronto_merged.loc[toronto_merged['Cluster Labels'] == 3, toronto_merged.columns[[1] + list(range(5, toronto_merged.shape[1]))]]\n\ntoronto_merged.loc[toronto_merged['Cluster Labels'] == 4, toronto_merged.columns[[1] + list(range(5, toronto_merged.shape[1]))]]\n\n# So we see the major differences: Neighborhoods in 0 have food/restaurants/businesses, in 1 they have playground/pool, in 2 they have parks as the most common, 3 its just one place that loves to drink, and 4 has baseball fields. \n","repo_name":"martjios/Coursera_Capstone","sub_path":"Week 3 Assignment - Course 9(3).ipynb","file_name":"Week 3 Assignment - Course 9(3).ipynb","file_ext":"py","file_size_in_byte":9767,"program_lang":"python","lang":"en","doc_type":"code","stars":0,"dataset":"github-jupyter-script","pt":"42"}
+{"seq_id":"5211881436","text":"# + [markdown] id=\"view-in-github\" colab_type=\"text\"\n# \n\n# + [markdown] id=\"3FCmKdMF6GW1\"\n# You want to minimize the cost of shipping goods from 2 different warehouses to 4 different customers. Each warehouse has a limited supply and each customer has a certain demand. We need to fulfil the demand of the customers by shipping products from given warehouses such that the overall cost of shipping is minimum and we are also able to satisfy the customer demands using limited supply available with each warehouse.\n# \n#\n# The customer demands and the warehouse availability is as follows.\n#\n# \n\n# + colab={\"base_uri\": \"https://localhost:8080/\"} id=\"R8PvKpH26GW7\" outputId=\"587d7102-02b8-4029-d302-74d174f63f5d\"\n# !pip install pulp\nfrom pulp import *\nimport pandas as pd\nimport numpy as np\n\n# + id=\"TlFN8UCI6GW9\"\nn_warehouses = 2\nn_customers = 4\n\n# Cost Matrix\ncost_matrix = np.array([[1, 3, 0.5, 4],\n [2.5, 5, 1.5, 2.5]])\n# Demand Matrix\ncust_demands = np.array([35000, 22000, 18000, 30000])\n\n# Supply Matrix\nwarehouse_supply = np.array([60000, 80000])\n\n# + id=\"xQttVqIY6GW9\"\nmodel = LpProblem(\"Supply-Demand-Problem\", LpMinimize)\n\n# + id=\"ryWhlYnN6GW-\" outputId=\"35cf709e-c73a-4562-c58c-d47ae25867ab\"\nvariable_names = [str(i)+str(j) for j in range(1, n_customers+1) for i in range(1, n_warehouses+1)]\nvariable_names.sort()\nprint(\"Variable Indices:\", variable_names)\n\n# + id=\"bv2mzrsy6GW-\" outputId=\"a83ed67d-0e85-4311-fade-692b591ebb45\"\nDV_variables = LpVariable.matrix(\"X\", variable_names, cat = \"Integer\", lowBound= 0 )\nallocation = np.array(DV_variables).reshape(2,4)\nprint(\"Decision Variable/Allocation Matrix: \")\nprint(allocation)\n\n# + id=\"ZmdGBbCm6GW_\" outputId=\"fb9b68ac-5a19-44ce-f795-04a6c9d3d7df\"\nobj_func = lpSum(allocation*cost_matrix)\nprint(obj_func)\nmodel += obj_func\nprint(model)\n\n# + id=\"2SKvwRK06GXA\" outputId=\"e11d39d3-6ad9-4fb6-c8a3-0cd43971a4a4\"\n#Supply Constraints\nfor i in range(n_warehouses):\n print(lpSum(allocation[i][j] for j in range(n_customers)) <= warehouse_supply[i])\n model += lpSum(allocation[i][j] for j in range(n_customers)) <= warehouse_supply[i] , \"Supply Constraints \" + str(i)\n\n# + id=\"PVCqeCmF6GXA\" outputId=\"c3044f6a-781a-4b4d-cdc8-94fcbee7a13a\"\n# Demand Constraints\nfor j in range(n_customers):\n print(lpSum(allocation[i][j] for i in range(n_warehouses)) >= cust_demands[j])\n model += lpSum(allocation[i][j] for i in range(n_warehouses)) >= cust_demands[j] , \"Demand Constraints \" + str(j)\n\n# + id=\"W9a0QVXN6GXB\" outputId=\"5c1e542d-cadb-450c-8452-18885abfcdf0\"\nprint(model)\n\n# + id=\"Fvrb9Flb6GXC\" outputId=\"28131e23-cafc-4ac3-e96a-b3cb075d251a\"\n#model.solve()\nmodel.solve(PULP_CBC_CMD())\n\nstatus = LpStatus[model.status]\n\nprint(status)\n\n# + id=\"_ZAS5KQm6GXC\" outputId=\"9995464f-e993-4fde-b0c0-0ff723341eba\"\nprint(\"Total Cost:\", model.objective.value())\n\n# Decision Variables\n\nfor v in model.variables():\n try:\n print(v.name,\"=\", v.value())\n except:\n print(\"error couldnt find value\")\n\n# + id=\"JpoEEc736GXC\" outputId=\"7c4c62fc-912a-42de-f1a4-e6864b0b636e\"\n# Warehouse 1 and Warehouse 2 required capacity\n\nfor i in range(n_warehouses):\n print(\"Warehouse \", str(i+1))\n print(lpSum(allocation[i][j].value() for j in range(n_customers)))\n","repo_name":"dele22922/tarea3_feat_eng","sub_path":"EjemploPulpTransp.ipynb","file_name":"EjemploPulpTransp.ipynb","file_ext":"py","file_size_in_byte":3531,"program_lang":"python","lang":"en","doc_type":"code","stars":0,"dataset":"github-jupyter-script","pt":"42"}
+{"seq_id":"26712007682","text":"import pandas as pd \nimport numpy as np\nimport matplotlib.pyplot as plt\n# %matplotlib inline \n\ndf = pd.read_csv('.\\Data\\Rios.csv',index_col=0)\ndf.head(5)\n\n# ## Histograma para ver como estan distribuidos nuestros valores \n\nplt.title('Rios') #Matplotlib\nplt.hist(df['Millas'], edgecolor='black', linewidth=1);\n\ndf['Millas'].hist(edgecolor='black', linewidth=1); #Pandas\n\n# ## Hacemos un grafico de caja para ver los outliers \n\nimport seaborn as sns\nsns.boxplot(df.Millas);\n\nnum_features = df[['Millas']] #Pandas\nnum_features.boxplot(grid=False);\n\ndf.boxplot(); #Pandas \n\nplt.boxplot(df['Millas']); #Matplotlib\n\n# ## Valores estadisticos de los Outliers\n\nQ1 = df['Millas'].quantile(0.25)\nQ3 = df['Millas'].quantile(0.75)\nRI = Q3 - Q1\nMediana = df['Millas'].median()\nMin = df['Millas'].min()\nMax = df['Millas'].max()\n\nprint(f'Primer Cuartil:{Q1}')\nprint(f'Segundo Cuartil:{Q3}')\nprint(f'Rango Intercuartil: {RI}')\nprint(f'Mediana:{Mediana}')\nprint(f'Min:{Min}')\nprint(f'Max:{Max}')\n\n# ## Calculo de limites\n\nlim_inf = Q1 - 1.5 * RI #Para los valores que aparecen en el box, lim inf [-245 o 135min] se queda con 135\nlim_sup = Q3 + 1.5 * RI #lim sup [1235 o 3710max] se queda con 1235\n[lim_inf,lim_sup]\n\n# ## Ubicación de los Outliers\n\nmask_outliers = ((df['Millas'] < lim_inf) | (df['Millas'] > lim_sup))\nmask_outliers\n\ndf_outliers = df[mask_outliers] #Me trae los outliers\ndf_outliers\n\ndf_outliers.sort_values('Millas',ascending=False)\n\n# ## Se sacan los datos ya sin outliers\n\nmask_outliers = ((df['Millas'] >= lim_inf) & (df['Millas'] <= lim_sup))\ndf_no_outliers = df[mask_outliers]\ndf_no_outliers\n\n# ## Se vuelve a graficar para corroborar que ya no hay outliers (Se repite el proceso hasta que se eliminan todos los outliers)\n\nplt.boxplot(df_no_outliers['Millas']);\n\nQ1 = df_no_outliers['Millas'].quantile(0.25)\nQ3 = df_no_outliers['Millas'].quantile(0.75)\nRI = Q3 - Q1\nMediana = df_no_outliers['Millas'].median()\nMin = df_no_outliers['Millas'].min()\nMax = df_no_outliers['Millas'].max()\n\nprint(f'Primer Cuartil:{Q1}')\nprint(f'Segundo Cuartil:{Q3}')\nprint(f'Rango Intercuartil: {RI}')\nprint(f'Mediana:{Mediana}')\nprint(f'Min:{Min}')\nprint(f'Max:{Max}')\n\nlim_inf = Q1 - 1.5 * RI #Para los valores que aparecen en el box, lim inf [-245 o 135min] se queda con 135\nlim_sup = Q3 + 1.5 * RI #lim sup [1235 o 3710max] se queda con 1235\n[lim_inf,lim_sup]\n\nmask = ((df_no_outliers['Millas'] >= lim_inf) & (df_no_outliers['Millas'] <= lim_sup))\ndf_no_outliers2 = df_no_outliers[mask]\n\nplt.boxplot(df_no_outliers2['Millas']);\n\nQ1 = df_no_outliers2['Millas'].quantile(0.25)\nQ3 = df_no_outliers2['Millas'].quantile(0.75)\nRI = Q3 - Q1\nMediana = df_no_outliers2['Millas'].median()\nMin = df_no_outliers2['Millas'].min()\nMax = df_no_outliers2['Millas'].max()\n\nlim_inf = Q1 - 1.5 * RI #Para los valores que aparecen en el box, lim inf [-245 o 135min] se queda con 135\nlim_sup = Q3 + 1.5 * RI #lim sup [1235 o 3710max] se queda con 1235\n[lim_inf,lim_sup]\n\nmask = ((df_no_outliers2['Millas'] >= lim_inf) & (df_no_outliers2['Millas'] <= lim_sup))\ndf_no_outliers3 = df_no_outliers2[mask]\n\nplt.boxplot(df_no_outliers3['Millas']);\n","repo_name":"rocooper7/Machine-Learning","sub_path":"5.Curso Prof ML/3. OutlierDetect&Remove/IQR/IQR2.ipynb","file_name":"IQR2.ipynb","file_ext":"py","file_size_in_byte":3198,"program_lang":"python","lang":"en","doc_type":"code","stars":0,"dataset":"github-jupyter-script","pt":"42"}
+{"seq_id":"33704977442","text":"# # Experiment One - R models\n\n# + tags=[\"parameters\"]\nfrom typing import Literal\n\n# decide which dataset and Cal, Tuning and Test set to use\nPREPROCESSING: Literal[\"raw\", \"anderson\", \"mishra\", \"mishra_adj\"] = \"anderson\"\nSET_SPLIT: Literal[\"anderson\", \"mishra\", \"mishra_with_outliers\"] = \"anderson\"\nTRAINING_ORDER: Literal[\"anderson\", \"mishra\", \"mishra_with_outliers\", \"random_1\", \"random_2\", \"random_3\"] = \"anderson\"\nINCLUDE_MISHRA_OUTLIERS: bool = True # True, False\n# -\n\n# ## 1.0 Setup\n\n# +\nimport datetime as dt\nimport os\nimport pandas as pd\n\nimport evaluation\n# -\n\nparams = {\n \"preprocessing\": PREPROCESSING,\n \"set_split\": SET_SPLIT,\n \"training_order\": TRAINING_ORDER,\n \"include_mishra_outliers\": INCLUDE_MISHRA_OUTLIERS,\n \"x_scaled\": False,\n}\n\n# ### 1.1 Prepare Dataset\n\n# +\n# prepare data based on set parameters above\n\n# read in data that has been preprocessed\nif PREPROCESSING == \"anderson\":\n data_file = \"data/input/anderson&etal_2020_pretreatment.csv\"\n # data_file = \"../../data/processed/NAnderson2020MendeleyMangoNIRData_Anderson2020pretreat_AllSetSplits.csv\"\nelif PREPROCESSING in (\"raw\", \"mishra\"):\n data_file = \"data/input/mishra&passos_2021_pretreatment.csv\"\n # data_file = \"../../data/processed/NAnderson2020MendeleyMangoNIRData_Mishra2021Aug_AllSetSplits.csv\"\nelif PREPROCESSING == \"mishra_adj\":\n data_file = \"data/input/mishra&passos_2021_pretreatment_adjusted.csv\"\n # data_file = \"../../data/processed/NAnderson2020MendeleyMangoNIRData_Mishra2021Aug2_AllSetSplits.csv\"\nelse:\n raise Exception(\"choose a valid preprocessing\")\ndata = pd.read_csv(data_file).drop(columns=[\"Unnamed: 0\"])\ndata.insert(0, \"sample_order\", 0)\ndata.insert(1, \"sample_no\", 0)\ndata.insert(2, \"set\", 0)\n\n# include or exclude mishra outliers\nif not INCLUDE_MISHRA_OUTLIERS:\n data = data.query(\"set_mishra != 'Outlier'\").copy()\n\n# assign set split\nif SET_SPLIT == \"anderson\":\n data[\"sample_no\"] = data[\"sample_no_anderson\"]\n data[\"set\"] = data[\"set_anderson\"]\nelif SET_SPLIT == \"mishra\" and not INCLUDE_MISHRA_OUTLIERS:\n data[\"sample_no\"] = data[\"sample_no_mishra\"]\n data[\"set\"] = data[\"set_mishra\"]\nelif SET_SPLIT == \"mishra_with_outliers\" and INCLUDE_MISHRA_OUTLIERS:\n data[\"sample_no\"] = data[\"sample_no_mishra_with_outliers\"]\n data[\"set\"] = data[\"set_mishra_with_outliers\"]\nelse:\n raise Exception(\"choose a valid set split\")\n\n# order data based on ordering selection\nif TRAINING_ORDER == \"mishra\" and SET_SPLIT == \"mishra_with_outliers\":\n TRAINING_ORDER = \"mishra_with_outliers\"\nelif TRAINING_ORDER == \"mishra_with_outliers\" and SET_SPLIT == \"mishra\":\n TRAINING_ORDER = \"mishra\"\ndata.sort_values(by=f\"sample_no_{TRAINING_ORDER}\", inplace=True)\ndata.reset_index(inplace=True, drop=True)\ndata[\"sample_order\"] = data.index + 1\n\ndata.groupby([\"category\", \"set\"]).size()\n# -\n\n# output to csv to be read by r script\ndata.to_csv(\"data/input_r/mango_meta&spectra.csv\")\n\n# ## 2.0 Build Models\n\n# ### 2.2 ANN Anderson et al. (2020)\n#\n# Model replicated in R, please see anderson-etal-2021-ann-model.R script for model training and predictions.\n#\n# Predicted values are read in below and tested.\n\n# +\npredictions_file = \"data/predictions/anderson_etal_2020_ann.csv\"\ntraining_time_file = \"data/input_r/training_time.csv\"\nprediction_time_file = \"data/input_r/prediction_time.csv\"\n\ny_test_comparison = pd.read_csv(predictions_file)\ntraining_time = pd.read_csv(training_time_file)\ntraining_time = training_time[\"x\"].iloc[0]\ntraining_time = dt.timedelta(minutes=training_time) if training_time < 8 else dt.timedelta(seconds=training_time)\nprediction_time = pd.read_csv(prediction_time_file)\nprediction_time = prediction_time[\"x\"].iloc[0]\nprediction_time = dt.timedelta(minutes=prediction_time) if prediction_time < 8 else dt.timedelta(seconds=prediction_time)\nos.remove(training_time_file)\nos.remove(prediction_time_file)\ny_test_comparison\n# -\n\nevaluation.calculate_metrics(\n y_true=y_test_comparison[\"actual\"],\n y_pred=y_test_comparison[\"predicted\"],\n)\n\n# statistical tests against best reported model (Mishra & Passos, 2021)\nmodel_name = \"mishra&passos_2021_best_cnn\"\nbase_model_preds = pd.read_csv(f\"data/predictions/{model_name}.csv\")\nevaluation.significance_difference(\n e_1=base_model_preds.query(\"test_data == 'test'\")[\"y_error\"].to_numpy(),\n e_2=(y_test_comparison[\"actual\"] - y_test_comparison[\"predicted\"]).to_numpy()\n)\n\ntraining_time\n\nprediction_time\n\nprediction_time/len(y_test_comparison)\n\n\n","repo_name":"jemjemwalsh/research-deep-learning-nirs","sub_path":"research/experiments/one/experiment_one_r_models.ipynb","file_name":"experiment_one_r_models.ipynb","file_ext":"py","file_size_in_byte":4461,"program_lang":"python","lang":"en","doc_type":"code","stars":1,"dataset":"github-jupyter-script","pt":"42"}
+{"seq_id":"18514383317","text":"import keras\nimport numpy as np\nfrom keras import layers\nimport random\nimport sys\n\npath = 'phr1.txt'\ntext = open(path, encoding='utf-8').read().lower()\nprint('Corpus length:', len(text))\n\n# +\nlines = text.splitlines()\nphrases, questions = [''], []\nfor line in lines:\n line = line.strip()\n if line:\n phrases[-1] += ' '+line\n if len(questions)\n# -\n\n# # リスト内包表記(教科書P.173)\n\n# ## リスト 参考サイト\n# ## [参考サイト1【Python早見表】リスト](https://chokkan.github.io/python/06list.html)\n#\n# ## [参考サイト2【Python入門】リスト](https://utokyo-ipp.github.io/2/2-2.html)\n#\n# ## [参考サイト3【Python公式】リスト型(list)](https://docs.python.org/ja/3/library/stdtypes.html?highlight=list#lists)\n\n# ## リスト内包表記とは\n# - listの中にfor inを書いて新しいリストを作る構文です\n# - Pythonではよく使用されます。Pythonの代表的な構文の１つです。\n#\n# ### 通常のfor文\n# ```python\n# a1 = []\n# for i in range(10):\n# a1.append(i)\n# ```\n#\n# ### リスト内包表記を使うと...？\n# ```python\n# a1 = [i for i in range(10)]\n# ```\n#\n# - このようにスッキリ書けます\n# - ソースコードを左から\"list of i for all i in the range of ten\"のように英語読みをすると解釈しやすいです。なので日本語読みす���ときは右から読んだほうがわかりやすいかもしれませんね。\n#\n# ### リスト内包表記の構文\n#\n# ```python\n# [式 for 変数 in イテラブル]\n# ```\n# - 構文的には in の場所に「イテラブル」という、値を列挙可能なオブジェクトを定義することになっています。\n#\n# ### イテラブル(iterable)とは？\n# - Pythonのイテラブル(iterable)オブジェクトは、リストの他に、文字列、range()で作るシーケンス、辞書(dict)、tuple、setなどがあります。\n#\n\n# ## サンプルプログラム10 リスト内包表記(教科書P.173)\n# - リスト内包表記でrangeを使ってリストを作る\n# - リスト内包表記でリストを加工して新たに作る\n# - zipを使って2つのリストを使って加工したリストを新たに作る\n# - リスト内包表記で文字列で単位を追加したリストを新たに作る\n\n# リストを作る\nincrements1 = [i for i in range(10)] # iterableにrangeを使って0〜9までのリストを作る\nincrements5 = [i*5 for i in increments1] # iterableにlistを使って5刻みの0〜45までのリストを作る\nprint('increments1 =',increments1)\nprint('increments5 =',increments5)\n\n# - range を使用して簡単に1行でリストを作ることが出来ます。\n# - listを使用して値を加工したリストを新たに作ることも出来ます。\n# - 取り出した要素をprint文で出力しています\n\n# zipを使って2つのリストを足し合わせたリストを新たに作る\nadd_list = [n1+n5 for n1,n5 in zip(increments1,increments5)]\nprint('add_list =',add_list)\n\n# 文字列で”cm”と単位をつけたリストを作る\nunit_cm = [f\"{i}cm\" for i in increments5]\nprint('unit_cm =',unit_cm)\n\n\n\n# ## 練習プログラム10 (0.8点)\n# - リスト内包表記でrangeを使ってリストを作る\n# - リスト内包表記でリストを加工して新たに作る\n# - zipを使って2つのリストを使って加工したリストを新たに作る\n# - リスト内包表記で文字列で単位を追加したリストを新たに作る\n\n# +\n# 1. リスト内包表記でrangeを使ってリストを作ってみましょう\n# 入れる値は何でも良いですが、このNotebookのサンプルプログラムや教科書の値とは違う独自の値を入れましょう\n\n\n\n# +\n# 2. リスト内包表記でリストを加工して新たに作ってみましょう\n\n\n\n# +\n# 3. zipを使って2つのリストを使って加工したリストを新たに作ってみましょう\n\n\n\n# +\n# 4. リスト内包表記で文字列で単位を追加したリストを新たに作ってみましょう\n\n\n# -\n\n\n\n# ## 条件付きリスト内包表記の構文\n# リスト内包表記にifを使って条件式を追加することで、条件にみあったリストを作ることが出来ます\n# ```python\n# [式 for 変数 in イテラブル if 条件式]\n# ```\n\n# ## サンプルプログラム11 条件付きのリスト内包表記(教科書P.175)\n# - 条件付きリスト内包表記\n# - 2つ以上の条件付きリスト内包表記\n\n# 条件付きリスト内包表記：1〜9のリストから２〜４の値だけ取り出してリストを作る\nnumber2to4 = [num for num in increments1 if 2<=num<=4]\nprint('number2to4 =',number2to4)\n\n# 2つ以上の条件付きリスト内包表記：２〜8の値で奇数だけ取り出してリストを作る\nodd2to8 = [num for num in increments1 if 2<=num<=8 if num%2==1]\nprint('odd2to8 =',odd2to8)\n\n# ## 練習プログラム11 (0.4点)\n# - 条件付きリスト内包表記を作る\n# - 2つ以上の条件付きリスト内包表記を作る\n\n# +\n# 1. 条件付きリスト内包表記を作ってみましょう\n\n\n\n# +\n# 2. リ2つ以上の条件付きリスト内包表記を使用して、新たにリストを作ってみましょう\n\n\n# -\n\n\n\n# ## for in を複数含める構文\n# リスト内包表記にfor inを複数使うことで、ネスト構造のデータを取り出したり、作り出すことができます\n# ```python\n# [式 for 変数1 in イテラブル1 for 変数2 in イテラブル2]\n# ```\n\n# ## サンプルプログラム12 for in を複数含めるリスト内包表記(教科書P.176) (※)\n# - for in を２つ使ったリスト内包表記で2次元配列を作る\n# - for in を２つ使ったリスト内包表記で2次元配列から1次元配列を作る\n\n# for in を２つ使って2次元配列を作る\n# 1feetをcmに直した値を入れる（1feet ≒ 30.48cm のため、cmのときは30.48倍した値を入れる）\nunit = [\"cm\", \"feet\"]\nunit_numbers1 = [[f\"{num}feet\" if unit1=='feet' else f\"{num*30.48}cm\" for unit1 in unit] for num in increments1]\nunit_numbers1\n\n# for in を２つ使って2次元配列から1次元配列を作る\nunit_numbers2 = [unit_num3 for unit_num2 in unit_numbers1 for unit_num3 in unit_num2]\nunit_numbers2\n\n# ## 練習プログラム12 (0.4点)\n# - for in を２つ使ったリスト内包表記で2次元配列を作ってみましょう\n# - for in を２つ使ったリスト内包表記で2次元配列から1次元配列を作ってみましょう\n\n# +\n# 1. for in を２つ使ったリスト内包表記で2次元配列を作ってみましょう\n\n\n\n# +\n# 2. for in を２つ使ったリスト内包表記で2次元配列から1次元配列を作ってみましょう\n\n\n# -\n\n\n\n\n\n\n\n\n","repo_name":"YasuharuSuzuki/23_programing1","sub_path":"09list/08リスト内包表記.ipynb","file_name":"08リスト内包表記.ipynb","file_ext":"py","file_size_in_byte":6694,"program_lang":"python","lang":"ja","doc_type":"code","stars":0,"dataset":"github-jupyter-script","pt":"41"}
+{"seq_id":"2771709710","text":"# ### Titan Bagus Bramantyo\n\n# Final Project to classify batik style using LBP (Local Binary Pattern) and KNN (K-Nearest Neighboor) algorithms.\n\n# #### ✅ Importing library\n\nimport pandas as pd\nimport numpy as np\nimport cv2 as cv\nimport matplotlib.pyplot as plt\nfrom sklearn.neighbors import KNeighborsClassifier\nfrom sklearn.naive_bayes import MultinomialNB\nfrom skimage.feature import local_binary_pattern\n\n# #### ✅ Config for LBP\n\n# +\nradius = 2\nn_points = 8 * radius\nMETHOD = 'uniform'\nplt.rcParams['font.size'] = 9\n\n# make function of LBP\ndef getdataTest(img):\n img_lbp = local_binary_pattern(img, n_points, radius, METHOD)\n img_lbp_hist,bins = np.histogram(img_lbp.ravel(),256,[0,256])\n img_lbp_hist = np.transpose(img_lbp_hist[0:18,np.newaxis])\n return img_lbp_hist\n\n\n# load image file\n## ---- ceplok style\ncp1_lbp = local_binary_pattern(cv.imread('dataset/training/ceplok-style/TC1.jpg', cv.IMREAD_GRAYSCALE), n_points, radius, METHOD)\ncp2_lbp = local_binary_pattern(cv.imread('dataset/training/ceplok-style/TC2.jpg', cv.IMREAD_GRAYSCALE), n_points, radius, METHOD)\ncp3_lbp = local_binary_pattern(cv.imread('dataset/training/ceplok-style/TC3.jpg', cv.IMREAD_GRAYSCALE), n_points, radius, METHOD)\ncp4_lbp = local_binary_pattern(cv.imread('dataset/training/ceplok-style/TC4.jpg', cv.IMREAD_GRAYSCALE), n_points, radius, METHOD)\ncp5_lbp = local_binary_pattern(cv.imread('dataset/training/ceplok-style/TC5.jpg', cv.IMREAD_GRAYSCALE), n_points, radius, METHOD)\ncp6_lbp = local_binary_pattern(cv.imread('dataset/training/ceplok-style/TC6.jpg', cv.IMREAD_GRAYSCALE), n_points, radius, METHOD)\ncp7_lbp = local_binary_pattern(cv.imread('dataset/training/ceplok-style/TC7.jpg', cv.IMREAD_GRAYSCALE), n_points, radius, METHOD)\ncp8_lbp = local_binary_pattern(cv.imread('dataset/training/ceplok-style/TC8.jpg', cv.IMREAD_GRAYSCALE), n_points, radius, METHOD)\ncp9_lbp = local_binary_pattern(cv.imread('dataset/training/ceplok-style/TC9.jpg', cv.IMREAD_GRAYSCALE), n_points, radius, METHOD)\ncp10_lbp = local_binary_pattern(cv.imread('dataset/training/ceplok-style/TC10.jpg', cv.IMREAD_GRAYSCALE), n_points, radius, METHOD)\ncp11_lbp = local_binary_pattern(cv.imread('dataset/training/ceplok-style/TC11.jpg', cv.IMREAD_GRAYSCALE), n_points, radius, METHOD)\ncp12_lbp = local_binary_pattern(cv.imread('dataset/training/ceplok-style/TC12.jpg', cv.IMREAD_GRAYSCALE), n_points, radius, METHOD)\ncp13_lbp = local_binary_pattern(cv.imread('dataset/training/ceplok-style/TC13.jpg', cv.IMREAD_GRAYSCALE), n_points, radius, METHOD)\ncp14_lbp = local_binary_pattern(cv.imread('dataset/training/ceplok-style/TC14.jpg', cv.IMREAD_GRAYSCALE), n_points, radius, METHOD)\ncp15_lbp = local_binary_pattern(cv.imread('dataset/training/ceplok-style/TC15.jpg', cv.IMREAD_GRAYSCALE), n_points, radius, METHOD)\ncp16_lbp = local_binary_pattern(cv.imread('dataset/training/ceplok-style/TC16.jpg', cv.IMREAD_GRAYSCALE), n_points, radius, METHOD)\ncp17_lbp = local_binary_pattern(cv.imread('dataset/training/ceplok-style/TC17.jpg', cv.IMREAD_GRAYSCALE), n_points, radius, METHOD)\ncp18_lbp = local_binary_pattern(cv.imread('dataset/training/ceplok-style/TC18.jpg', cv.IMREAD_GRAYSCALE), n_points, radius, METHOD)\ncp19_lbp = local_binary_pattern(cv.imread('dataset/training/ceplok-style/TC19.jpg', cv.IMREAD_GRAYSCALE), n_points, radius, METHOD)\ncp20_lbp = local_binary_pattern(cv.imread('dataset/training/ceplok-style/TC20.jpg', cv.IMREAD_GRAYSCALE), n_points, radius, METHOD)\ncp21_lbp = local_binary_pattern(cv.imread('dataset/training/ceplok-style/TC21.jpg', cv.IMREAD_GRAYSCALE), n_points, radius, METHOD)\ncp22_lbp = local_binary_pattern(cv.imread('dataset/training/ceplok-style/TC22.jpg', cv.IMREAD_GRAYSCALE), n_points, radius, METHOD)\ncp23_lbp = local_binary_pattern(cv.imread('dataset/training/ceplok-style/TC23.jpg', cv.IMREAD_GRAYSCALE), n_points, radius, METHOD)\ncp24_lbp = local_binary_pattern(cv.imread('dataset/training/ceplok-style/TC24.jpg', cv.IMREAD_GRAYSCALE), n_points, radius, METHOD)\ncp25_lbp = local_binary_pattern(cv.imread('dataset/training/ceplok-style/TC25.jpg', cv.IMREAD_GRAYSCALE), n_points, radius, METHOD)\ncp26_lbp = local_binary_pattern(cv.imread('dataset/training/ceplok-style/TC26.jpg', cv.IMREAD_GRAYSCALE), n_points, radius, METHOD)\ncp27_lbp = local_binary_pattern(cv.imread('dataset/training/ceplok-style/TC27.jpg', cv.IMREAD_GRAYSCALE), n_points, radius, METHOD)\ncp28_lbp = local_binary_pattern(cv.imread('dataset/training/ceplok-style/TC28.jpg', cv.IMREAD_GRAYSCALE), n_points, radius, METHOD)\ncp29_lbp = local_binary_pattern(cv.imread('dataset/training/ceplok-style/TC29.jpg', cv.IMREAD_GRAYSCALE), n_points, radius, METHOD)\ncp30_lbp = local_binary_pattern(cv.imread('dataset/training/ceplok-style/TC30.jpg', cv.IMREAD_GRAYSCALE), n_points, radius, METHOD)\n\n## ---- parang style\npr1_lbp = local_binary_pattern(cv.imread('dataset/training/parang-style/TP1.jpg', cv.IMREAD_GRAYSCALE), n_points, radius, METHOD)\npr2_lbp = local_binary_pattern(cv.imread('dataset/training/parang-style/TP2.jpg', cv.IMREAD_GRAYSCALE), n_points, radius, METHOD)\npr3_lbp = local_binary_pattern(cv.imread('dataset/training/parang-style/TP3.jpg', cv.IMREAD_GRAYSCALE), n_points, radius, METHOD)\npr4_lbp = local_binary_pattern(cv.imread('dataset/training/parang-style/TP4.jpg', cv.IMREAD_GRAYSCALE), n_points, radius, METHOD)\npr5_lbp = local_binary_pattern(cv.imread('dataset/training/parang-style/TP5.jpg', cv.IMREAD_GRAYSCALE), n_points, radius, METHOD)\npr6_lbp = local_binary_pattern(cv.imread('dataset/training/parang-style/TP6.jpg', cv.IMREAD_GRAYSCALE), n_points, radius, METHOD)\npr7_lbp = local_binary_pattern(cv.imread('dataset/training/parang-style/TP7.jpg', cv.IMREAD_GRAYSCALE), n_points, radius, METHOD)\npr8_lbp = local_binary_pattern(cv.imread('dataset/training/parang-style/TP8.jpg', cv.IMREAD_GRAYSCALE), n_points, radius, METHOD)\npr9_lbp = local_binary_pattern(cv.imread('dataset/training/parang-style/TP9.jpg', cv.IMREAD_GRAYSCALE), n_points, radius, METHOD)\npr10_lbp = local_binary_pattern(cv.imread('dataset/training/parang-style/TP10.jpg', cv.IMREAD_GRAYSCALE), n_points, radius, METHOD)\npr11_lbp = local_binary_pattern(cv.imread('dataset/training/parang-style/TP11.jpg', cv.IMREAD_GRAYSCALE), n_points, radius, METHOD)\npr12_lbp = local_binary_pattern(cv.imread('dataset/training/parang-style/TP12.jpg', cv.IMREAD_GRAYSCALE), n_points, radius, METHOD)\npr13_lbp = local_binary_pattern(cv.imread('dataset/training/parang-style/TP13.jpg', cv.IMREAD_GRAYSCALE), n_points, radius, METHOD)\npr14_lbp = local_binary_pattern(cv.imread('dataset/training/parang-style/TP14.jpg', cv.IMREAD_GRAYSCALE), n_points, radius, METHOD)\npr15_lbp = local_binary_pattern(cv.imread('dataset/training/parang-style/TP15.jpg', cv.IMREAD_GRAYSCALE), n_points, radius, METHOD)\npr16_lbp = local_binary_pattern(cv.imread('dataset/training/parang-style/TP16.jpg', cv.IMREAD_GRAYSCALE), n_points, radius, METHOD)\npr17_lbp = local_binary_pattern(cv.imread('dataset/training/parang-style/TP17.jpg', cv.IMREAD_GRAYSCALE), n_points, radius, METHOD)\npr18_lbp = local_binary_pattern(cv.imread('dataset/training/parang-style/TP18.jpg', cv.IMREAD_GRAYSCALE), n_points, radius, METHOD)\npr19_lbp = local_binary_pattern(cv.imread('dataset/training/parang-style/TP19.jpg', cv.IMREAD_GRAYSCALE), n_points, radius, METHOD)\npr20_lbp = local_binary_pattern(cv.imread('dataset/training/parang-style/TP20.jpg', cv.IMREAD_GRAYSCALE), n_points, radius, METHOD)\npr21_lbp = local_binary_pattern(cv.imread('dataset/training/parang-style/TP21.jpg', cv.IMREAD_GRAYSCALE), n_points, radius, METHOD)\npr22_lbp = local_binary_pattern(cv.imread('dataset/training/parang-style/TP22.jpg', cv.IMREAD_GRAYSCALE), n_points, radius, METHOD)\npr23_lbp = local_binary_pattern(cv.imread('dataset/training/parang-style/TP23.jpg', cv.IMREAD_GRAYSCALE), n_points, radius, METHOD)\npr24_lbp = local_binary_pattern(cv.imread('dataset/training/parang-style/TP24.jpg', cv.IMREAD_GRAYSCALE), n_points, radius, METHOD)\npr25_lbp = local_binary_pattern(cv.imread('dataset/training/parang-style/TP25.jpg', cv.IMREAD_GRAYSCALE), n_points, radius, METHOD)\npr26_lbp = local_binary_pattern(cv.imread('dataset/training/parang-style/TP26.jpg', cv.IMREAD_GRAYSCALE), n_points, radius, METHOD)\npr27_lbp = local_binary_pattern(cv.imread('dataset/training/parang-style/TP27.jpg', cv.IMREAD_GRAYSCALE), n_points, radius, METHOD)\npr28_lbp = local_binary_pattern(cv.imread('dataset/training/parang-style/TP28.jpg', cv.IMREAD_GRAYSCALE), n_points, radius, METHOD)\npr29_lbp = local_binary_pattern(cv.imread('dataset/training/parang-style/TP29.jpg', cv.IMREAD_GRAYSCALE), n_points, radius, METHOD)\npr30_lbp = local_binary_pattern(cv.imread('dataset/training/parang-style/TP30.jpg', cv.IMREAD_GRAYSCALE), n_points, radius, METHOD)\n\n## -- ceplok style\ncp1_lbp_hist,bins = np.histogram(cp1_lbp.ravel(),256,[0,256])\ncp2_lbp_hist,bins = np.histogram(cp2_lbp.ravel(),256,[0,256])\ncp3_lbp_hist,bins = np.histogram(cp3_lbp.ravel(),256,[0,256])\ncp4_lbp_hist,bins = np.histogram(cp4_lbp.ravel(),256,[0,256])\ncp5_lbp_hist,bins = np.histogram(cp5_lbp.ravel(),256,[0,256])\ncp6_lbp_hist,bins = np.histogram(cp6_lbp.ravel(),256,[0,256])\ncp7_lbp_hist,bins = np.histogram(cp7_lbp.ravel(),256,[0,256])\ncp8_lbp_hist,bins = np.histogram(cp8_lbp.ravel(),256,[0,256])\ncp9_lbp_hist,bins = np.histogram(cp9_lbp.ravel(),256,[0,256])\ncp10_lbp_hist,bins = np.histogram(cp10_lbp.ravel(),256,[0,256])\ncp11_lbp_hist,bins = np.histogram(cp11_lbp.ravel(),256,[0,256])\ncp12_lbp_hist,bins = np.histogram(cp12_lbp.ravel(),256,[0,256])\ncp13_lbp_hist,bins = np.histogram(cp13_lbp.ravel(),256,[0,256])\ncp14_lbp_hist,bins = np.histogram(cp14_lbp.ravel(),256,[0,256])\ncp15_lbp_hist,bins = np.histogram(cp15_lbp.ravel(),256,[0,256])\ncp16_lbp_hist,bins = np.histogram(cp16_lbp.ravel(),256,[0,256])\ncp17_lbp_hist,bins = np.histogram(cp17_lbp.ravel(),256,[0,256])\ncp18_lbp_hist,bins = np.histogram(cp18_lbp.ravel(),256,[0,256])\ncp19_lbp_hist,bins = np.histogram(cp19_lbp.ravel(),256,[0,256])\ncp20_lbp_hist,bins = np.histogram(cp20_lbp.ravel(),256,[0,256])\ncp21_lbp_hist,bins = np.histogram(cp21_lbp.ravel(),256,[0,256])\ncp22_lbp_hist,bins = np.histogram(cp22_lbp.ravel(),256,[0,256])\ncp23_lbp_hist,bins = np.histogram(cp23_lbp.ravel(),256,[0,256])\ncp24_lbp_hist,bins = np.histogram(cp24_lbp.ravel(),256,[0,256])\ncp25_lbp_hist,bins = np.histogram(cp25_lbp.ravel(),256,[0,256])\ncp26_lbp_hist,bins = np.histogram(cp26_lbp.ravel(),256,[0,256])\ncp27_lbp_hist,bins = np.histogram(cp27_lbp.ravel(),256,[0,256])\ncp28_lbp_hist,bins = np.histogram(cp28_lbp.ravel(),256,[0,256])\ncp29_lbp_hist,bins = np.histogram(cp29_lbp.ravel(),256,[0,256])\ncp30_lbp_hist,bins = np.histogram(cp30_lbp.ravel(),256,[0,256])\n\n## ---- parang style\npr1_lbp_hist,bins = np.histogram(pr1_lbp.ravel(),256,[0,256])\npr2_lbp_hist,bins = np.histogram(pr2_lbp.ravel(),256,[0,256])\npr3_lbp_hist,bins = np.histogram(pr3_lbp.ravel(),256,[0,256])\npr4_lbp_hist,bins = np.histogram(pr4_lbp.ravel(),256,[0,256])\npr5_lbp_hist,bins = np.histogram(pr5_lbp.ravel(),256,[0,256])\npr6_lbp_hist,bins = np.histogram(pr6_lbp.ravel(),256,[0,256])\npr7_lbp_hist,bins = np.histogram(pr7_lbp.ravel(),256,[0,256])\npr8_lbp_hist,bins = np.histogram(pr8_lbp.ravel(),256,[0,256])\npr9_lbp_hist,bins = np.histogram(pr9_lbp.ravel(),256,[0,256])\npr10_lbp_hist,bins = np.histogram(pr10_lbp.ravel(),256,[0,256])\npr11_lbp_hist,bins = np.histogram(pr11_lbp.ravel(),256,[0,256])\npr12_lbp_hist,bins = np.histogram(pr12_lbp.ravel(),256,[0,256])\npr13_lbp_hist,bins = np.histogram(pr13_lbp.ravel(),256,[0,256])\npr14_lbp_hist,bins = np.histogram(pr14_lbp.ravel(),256,[0,256])\npr15_lbp_hist,bins = np.histogram(pr15_lbp.ravel(),256,[0,256])\npr16_lbp_hist,bins = np.histogram(pr16_lbp.ravel(),256,[0,256])\npr17_lbp_hist,bins = np.histogram(pr17_lbp.ravel(),256,[0,256])\npr18_lbp_hist,bins = np.histogram(pr18_lbp.ravel(),256,[0,256])\npr19_lbp_hist,bins = np.histogram(pr19_lbp.ravel(),256,[0,256])\npr20_lbp_hist,bins = np.histogram(pr20_lbp.ravel(),256,[0,256])\npr21_lbp_hist,bins = np.histogram(pr21_lbp.ravel(),256,[0,256])\npr22_lbp_hist,bins = np.histogram(pr22_lbp.ravel(),256,[0,256])\npr23_lbp_hist,bins = np.histogram(pr23_lbp.ravel(),256,[0,256])\npr24_lbp_hist,bins = np.histogram(pr24_lbp.ravel(),256,[0,256])\npr25_lbp_hist,bins = np.histogram(pr25_lbp.ravel(),256,[0,256])\npr26_lbp_hist,bins = np.histogram(pr26_lbp.ravel(),256,[0,256])\npr27_lbp_hist,bins = np.histogram(pr27_lbp.ravel(),256,[0,256])\npr28_lbp_hist,bins = np.histogram(pr28_lbp.ravel(),256,[0,256])\npr29_lbp_hist,bins = np.histogram(pr29_lbp.ravel(),256,[0,256])\npr30_lbp_hist,bins = np.histogram(pr30_lbp.ravel(),256,[0,256])\n\n# transpose from LBP -- ceplok style\ncp1_lbp_hist=np.transpose(cp1_lbp_hist[0:18,np.newaxis])\ncp2_lbp_hist=np.transpose(cp2_lbp_hist[0:18,np.newaxis])\ncp3_lbp_hist=np.transpose(cp3_lbp_hist[0:18,np.newaxis])\ncp4_lbp_hist=np.transpose(cp4_lbp_hist[0:18,np.newaxis])\ncp5_lbp_hist=np.transpose(cp5_lbp_hist[0:18,np.newaxis])\ncp6_lbp_hist=np.transpose(cp6_lbp_hist[0:18,np.newaxis])\ncp7_lbp_hist=np.transpose(cp7_lbp_hist[0:18,np.newaxis])\ncp8_lbp_hist=np.transpose(cp8_lbp_hist[0:18,np.newaxis])\ncp9_lbp_hist=np.transpose(cp9_lbp_hist[0:18,np.newaxis])\ncp10_lbp_hist=np.transpose(cp10_lbp_hist[0:18,np.newaxis])\ncp11_lbp_hist=np.transpose(cp11_lbp_hist[0:18,np.newaxis])\ncp12_lbp_hist=np.transpose(cp12_lbp_hist[0:18,np.newaxis])\ncp13_lbp_hist=np.transpose(cp13_lbp_hist[0:18,np.newaxis])\ncp14_lbp_hist=np.transpose(cp14_lbp_hist[0:18,np.newaxis])\ncp15_lbp_hist=np.transpose(cp15_lbp_hist[0:18,np.newaxis])\ncp16_lbp_hist=np.transpose(cp16_lbp_hist[0:18,np.newaxis])\ncp17_lbp_hist=np.transpose(cp17_lbp_hist[0:18,np.newaxis])\ncp18_lbp_hist=np.transpose(cp18_lbp_hist[0:18,np.newaxis])\ncp19_lbp_hist=np.transpose(cp19_lbp_hist[0:18,np.newaxis])\ncp20_lbp_hist=np.transpose(cp20_lbp_hist[0:18,np.newaxis])\ncp21_lbp_hist=np.transpose(cp21_lbp_hist[0:18,np.newaxis])\ncp22_lbp_hist=np.transpose(cp22_lbp_hist[0:18,np.newaxis])\ncp23_lbp_hist=np.transpose(cp23_lbp_hist[0:18,np.newaxis])\ncp24_lbp_hist=np.transpose(cp24_lbp_hist[0:18,np.newaxis])\ncp25_lbp_hist=np.transpose(cp25_lbp_hist[0:18,np.newaxis])\ncp26_lbp_hist=np.transpose(cp26_lbp_hist[0:18,np.newaxis])\ncp27_lbp_hist=np.transpose(cp27_lbp_hist[0:18,np.newaxis])\ncp28_lbp_hist=np.transpose(cp28_lbp_hist[0:18,np.newaxis])\ncp29_lbp_hist=np.transpose(cp29_lbp_hist[0:18,np.newaxis])\ncp30_lbp_hist=np.transpose(cp30_lbp_hist[0:18,np.newaxis])\n\n# transpose from LBP -- parang style\npr1_lbp_hist=np.transpose(pr1_lbp_hist[0:18,np.newaxis])\npr2_lbp_hist=np.transpose(pr2_lbp_hist[0:18,np.newaxis])\npr3_lbp_hist=np.transpose(pr3_lbp_hist[0:18,np.newaxis])\npr4_lbp_hist=np.transpose(pr4_lbp_hist[0:18,np.newaxis])\npr5_lbp_hist=np.transpose(pr5_lbp_hist[0:18,np.newaxis])\npr6_lbp_hist=np.transpose(pr6_lbp_hist[0:18,np.newaxis])\npr7_lbp_hist=np.transpose(pr7_lbp_hist[0:18,np.newaxis])\npr8_lbp_hist=np.transpose(pr8_lbp_hist[0:18,np.newaxis])\npr9_lbp_hist=np.transpose(pr9_lbp_hist[0:18,np.newaxis])\npr10_lbp_hist=np.transpose(pr10_lbp_hist[0:18,np.newaxis])\npr11_lbp_hist=np.transpose(pr11_lbp_hist[0:18,np.newaxis])\npr12_lbp_hist=np.transpose(pr12_lbp_hist[0:18,np.newaxis])\npr13_lbp_hist=np.transpose(pr13_lbp_hist[0:18,np.newaxis])\npr14_lbp_hist=np.transpose(pr14_lbp_hist[0:18,np.newaxis])\npr15_lbp_hist=np.transpose(pr15_lbp_hist[0:18,np.newaxis])\npr16_lbp_hist=np.transpose(pr16_lbp_hist[0:18,np.newaxis])\npr17_lbp_hist=np.transpose(pr17_lbp_hist[0:18,np.newaxis])\npr18_lbp_hist=np.transpose(pr18_lbp_hist[0:18,np.newaxis])\npr19_lbp_hist=np.transpose(pr19_lbp_hist[0:18,np.newaxis])\npr20_lbp_hist=np.transpose(pr20_lbp_hist[0:18,np.newaxis])\npr21_lbp_hist=np.transpose(pr21_lbp_hist[0:18,np.newaxis])\npr22_lbp_hist=np.transpose(pr22_lbp_hist[0:18,np.newaxis])\npr23_lbp_hist=np.transpose(pr23_lbp_hist[0:18,np.newaxis])\npr24_lbp_hist=np.transpose(pr24_lbp_hist[0:18,np.newaxis])\npr25_lbp_hist=np.transpose(pr25_lbp_hist[0:18,np.newaxis])\npr26_lbp_hist=np.transpose(pr26_lbp_hist[0:18,np.newaxis])\npr27_lbp_hist=np.transpose(pr27_lbp_hist[0:18,np.newaxis])\npr28_lbp_hist=np.transpose(pr28_lbp_hist[0:18,np.newaxis])\npr29_lbp_hist=np.transpose(pr29_lbp_hist[0:18,np.newaxis])\npr30_lbp_hist=np.transpose(pr30_lbp_hist[0:18,np.newaxis])\n\n# -\n\n# #### ✅ Joining ceplok and parang dataset in one matric\n\n# data train\ndata_train = np.concatenate((\n cp1_lbp_hist,\n cp2_lbp_hist,\n cp3_lbp_hist,\n cp4_lbp_hist,\n cp5_lbp_hist,\n cp6_lbp_hist,\n cp7_lbp_hist,\n cp8_lbp_hist,\n cp9_lbp_hist,\n cp10_lbp_hist,\n cp11_lbp_hist,\n cp12_lbp_hist,\n cp13_lbp_hist,\n cp14_lbp_hist,\n cp15_lbp_hist,\n cp16_lbp_hist,\n cp17_lbp_hist,\n cp18_lbp_hist,\n cp19_lbp_hist,\n cp20_lbp_hist,\n cp21_lbp_hist,\n cp22_lbp_hist,\n cp23_lbp_hist,\n cp24_lbp_hist,\n cp25_lbp_hist,\n cp26_lbp_hist,\n cp27_lbp_hist,\n cp28_lbp_hist,\n cp29_lbp_hist,\n cp30_lbp_hist,\n pr1_lbp_hist,\n pr2_lbp_hist,\n pr3_lbp_hist,\n pr4_lbp_hist,\n pr5_lbp_hist,\n pr6_lbp_hist,\n pr7_lbp_hist,\n pr8_lbp_hist,\n pr9_lbp_hist,\n pr10_lbp_hist,\n pr11_lbp_hist,\n pr12_lbp_hist,\n pr13_lbp_hist,\n pr14_lbp_hist,\n pr15_lbp_hist,\n pr16_lbp_hist,\n pr17_lbp_hist,\n pr18_lbp_hist,\n pr19_lbp_hist,\n pr20_lbp_hist,\n pr21_lbp_hist,\n pr22_lbp_hist,\n pr23_lbp_hist,\n pr24_lbp_hist,\n pr25_lbp_hist,\n pr26_lbp_hist,\n pr27_lbp_hist,\n pr28_lbp_hist,\n pr29_lbp_hist,\n pr30_lbp_hist\n ), axis=0).astype(np.float32)\n\n# #### ✅ Label for data train\n\n# label for dataset\nlabel = [\"Batik Ceplok\",\"Batik Ceplok\",\"Batik Ceplok\",\"Batik Ceplok\",\"Batik Ceplok\",\t\n\t\"Batik Ceplok\",\"Batik Ceplok\",\"Batik Ceplok\",\"Batik Ceplok\",\"Batik Ceplok\",\n\t\"Batik Ceplok\",\"Batik Ceplok\",\"Batik Ceplok\",\"Batik Ceplok\",\"Batik Ceplok\",\n\t\"Batik Ceplok\",\"Batik Ceplok\",\"Batik Ceplok\",\"Batik Ceplok\",\"Batik Ceplok\",\n\t\"Batik Ceplok\",\"Batik Ceplok\",\"Batik Ceplok\",\"Batik Ceplok\",\"Batik Ceplok\",\n\t\"Batik Ceplok\",\"Batik Ceplok\",\"Batik Ceplok\",\"Batik Ceplok\",\"Batik Ceplok\",\n\t\"Batik Parang\",\"Batik Parang\",\"Batik Parang\",\"Batik Parang\",\"Batik Parang\",\n\t\"Batik Parang\",\"Batik Parang\",\"Batik Parang\",\"Batik Parang\",\"Batik Parang\",\n\t\"Batik Parang\",\"Batik Parang\",\"Batik Parang\",\"Batik Parang\",\"Batik Parang\",\n\t\"Batik Parang\",\"Batik Parang\",\"Batik Parang\",\"Batik Parang\",\"Batik Parang\",\n\t\"Batik Parang\",\"Batik Parang\",\"Batik Parang\",\"Batik Parang\",\"Batik Parang\",\n\t\"Batik Parang\",\"Batik Parang\",\"Batik Parang\",\"Batik Parang\",\"Batik Parang\"]\n\n# #### ✅ Build KNN model\n\n# +\nfrom sklearn.model_selection import cross_val_score\n\n# KNN Modeling\nfor i in range (2,25):\n knn = KNeighborsClassifier(n_neighbors=i)\n acc_knn = cross_val_score(knn, data_train, label, cv=5, scoring='accuracy')\n print(\"Accuracy result using K-Nearest Neighbor K =\",i,\" : \", acc_knn.mean())\n\nknn = KNeighborsClassifier(n_neighbors=4)\nknn.fit(data_train,label)\n# -\n\n# #### ✅ Build SVM model\n\n# +\nfrom sklearn.svm import SVC\n\nsvc = SVC(kernel='linear', C=1)\nsvc.fit(data_train, label)\n\nacc_svc = cross_val_score(svc, data_train, label, cv=5, scoring='accuracy')\nprint(\"Accuracy result using SVM : \", acc_svc.mean())\n# -\n\n# #### ✅ Build Naive Bayes model\n\n# +\nnb = MultinomialNB()\n\nacc_nb = cross_val_score(nb, data_train, label, cv=5, scoring='accuracy')\nnb.fit(data_train, label)\n\nprint(\"Accuracy result using SVM : \", acc_svc.mean())\n# -\n\n# #### ✅ Detection testing\n\n# +\ngambar = cv.imread(\"dataset/testing/ceplok_style/SC012.jpg\", cv.IMREAD_GRAYSCALE)\n\nx = getdataTest(gambar)\nknn_pred = knn.predict(x)\nsvm_pred = svc.predict(x)\nnb_pred = nb.predict(x)\n\nprint(\"Detection KNN result : \", knn_pred)\nprint(\"Detection SVM result : \", svm_pred)\nprint(\"Detection Naive Bayes result : \", nb_pred)\n\ncv.namedWindow('Identified image', cv.WINDOW_NORMAL)\ncv.imshow('Identified image' , gambar)\ncv.waitKey(0)\ncv.destroyAllWindows()\n","repo_name":"katibpasha/data_science_portfolio","sub_path":"Project/Machine-Learning/Rekognisi-corak-batik/CODE - Batik Style Recognition.ipynb","file_name":"CODE - Batik Style Recognition.ipynb","file_ext":"py","file_size_in_byte":21244,"program_lang":"python","lang":"en","doc_type":"code","stars":2,"dataset":"github-jupyter-script","pt":"41"}
+{"seq_id":"21489913218","text":"# # Data Exploration\n\n# ## Importing necessary Libraries and the data set.\n\n# Pour connaitre le dossier que Jupyter utilise pour acceder aux fichiers (datasets)\n#\n# # !echo %cd%\n#\n# # !dir\n\nimport numpy as np\nimport matplotlib.pyplot as plt\nimport pandas as pd\nimport seaborn as sb\n\n\ndataset = pd.read_csv(\"insurance.csv\")\n\n# ## Understanding the variables in the data set.\n\ndataset.head()\n\n# ## Now Check the dimension of the given data set.\n\ndataset.shape\n\n# ## Dropping Duplicates rows in the columns if any\n\ndataset = dataset.drop_duplicates()\ndataset.shape\n\n# +\n# dataset[dataset.isin(list(dataset.drop_duplicates()))].dropna()\n# -\n\n# we observe that there is one duplicate column in the data set.\n\n# ## Using describe method\n#\n# to have information about the variables\n\ndataset.describe()\n\n# ## Using value_counts() method\n#\n# on region and sex to identify the count of each category in that column.\n\ndataset['region'].value_counts()\n\ndataset['sex'].value_counts()\n\n# ## Creating Dummies\n\n# +\n#\tage\tsex\tbmi\tchildren\tsmoker_yes\tsmoker_no\tregion\tcharges\n#0\t19\tfemale\t27.900\t0\t1 0\tsouthwest\t16884.92400\n#1\t18\tmale\t33.770\t1\t0 1\tsoutheast\t1725.55230\n#2\t28\tmale\t33.000\t3\t0 1\tsoutheast\t4449.46200\n#3\t33\tmale\t22.705\t0\t0 1\tnorthwest\t21984.47061\n#4\t32\tmale\t28.880\t0\t0 1\tnorthwest\t3866.85520\n# -\n\nold_dataset = dataset\ndataset.shape\n\n# ### Scikit Learn Way\n\n# +\nfrom sklearn.preprocessing import LabelEncoder\n\nx = old_dataset.iloc[:,:-1].values\ny = old_dataset.iloc[:,6].values\n\nprint(x.shape)\nprint(x[:10])\n\nlabelencoder=LabelEncoder()\nx[:,1]=labelencoder.fit_transform(x[:,1]) \nx[:,4]=labelencoder.fit_transform(x[:,4])\nx[:,5]=labelencoder.fit_transform(x[:,5])\n\nprint(x[:10])\n# -\n\n# ### Pandas way\n\ndataset = pd.get_dummies(dataset, columns=['smoker', 'sex', 'region'])\ndataset.shape\n\nold_dataset.shape\n\ndataset.head()\n\n# +\n# dataset.head\n# -\n\n# ## Using pairplot method\n#\n# to understand the distribution of each variable with other\n\nsb.histplot(dataset['age'], kde=True)\n\n# sb.scatterplot(dataset, x='age', y='charges')\nsb.scatterplot(x=dataset['age'], y=dataset['charges'])\n\nsb.pairplot(dataset)\n\n# ## Plotting heat map\n#\n# to understand the correlation of variables\n\nsb.scatterplot(x=dataset['age'], y=dataset['age'])\n\n# +\n# sb.scatterplot(x=dataset['region_northwest'], y=dataset['smoker_no'], x_jitter=True, y_jitter=True)\n# -\n\ndataset.corr()\n\nsb.heatmap(dataset.corr(),annot=True,cmap='magma')\n\n# ## Understanding the distribution\n\n# ### of the charges\n\n# !pip install scipy\n\n# +\nfrom scipy.stats import skew\n\nskew(dataset['charges'])\n# -\n\n# \n\n# \n\nsb.histplot(dataset['charges'], kde=True)\n\n# ### of the age\n\nsb.histplot(dataset['age'], kde=True)\n\n# # Feature Exploration\n\n# ## How many females and males are smokers\n\nold_dataset[\"sex\"].head()\n\n#how many females and males are smokers\nsb.countplot(x = \"sex\", data = old_dataset)\n\n# ## Smoker who are below or equal to 18 years\n\n# Smoker who are below or eaual to 18\nsb.countplot(x='sex',data=old_dataset[(old_dataset.age <= 18)])\n\n# ## Number of smokers from different regions\n\nold_dataset.smoker=='yes'\n\n(old_dataset.sex=='female')\n\n# +\n# Si 0 -> Faux 0\n# Si != 0 -> Vrai 1\n\n# Val_1 Val_2 & (ET BINAIRE)\n# 0 0 0 \n# 1 0 0\n# 0 1 0\n# 1 1 1\n\n# | s'appelle un pipe\n# Val_1 Val_2 | (OU BINAIRE)\n# 0 0 0 \n# 1 0 1\n# 0 1 1\n# 1 1 1\n\n# ^ s'apelle 'OU EXCLUSIF' (XOR)\n# -\n\nold_dataset[(old_dataset.age<=25) & (old_dataset.smoker=='yes')]\n\n#Smokers from different regions\nsb.countplot(x='region',data=old_dataset)\n\n#Smokers from different regions\nsb.countplot(x='region',data=old_dataset[(old_dataset.smoker=='yes')])\n\n# ## Number of children\n\n# +\n#No. of childrens our patients have\n\nsb.countplot(x='children',data=dataset)\n# -\n\n# ## Number of children of smokers patients\n\n# +\n#No. of childrens of smokers patients \n\nsb.countplot(x='children',data=old_dataset[(old_dataset.smoker=='yes')])\n\n# +\n#No. of childrens of smokers patients \n\nsb.countplot(x='children',data=old_dataset[(old_dataset.smoker=='no')])\n# -\n\nold_dataset.columns\n\n# ## Charges in function of target\n\nsb.scatterplot(data=old_dataset, x=\"age\", y=\"charges\", hue=\"smoker\")\n\nsubset_wealthy = old_dataset[(old_dataset.charges>=30_000)]\n\nsb.scatterplot(data=subset_wealthy, x=\"age\", y=\"charges\", hue=\"smoker\")\n\nsubset_wealthy\n\n# ## Smokers on the basis of age\n\n#Smokers on the basis of age\nplt.figure(figsize=(14,7))\nsb.countplot(x='age',data=dataset[(dataset.smoker_yes==1)])\nplt.grid()\n\n# ## Non-Smokers on the basis of the age\n\n#non-Smokers on the basis of the age\nplt.figure(figsize=(14,7))\nsb.countplot(x='age',data=old_dataset[(old_dataset.smoker=='no')])\nplt.grid()\n\n# # Applying regression model\n\n# ## Splitting the data set into Features et Target\n\ndataset.head()\n\nX=dataset.drop(['smoker_no', 'smoker_yes'], axis=1).values\ny=dataset['smoker_yes'].values\nprint(X.shape, y.shape)\n\n# ## Splitting the data set into training and testing data set \n\nfrom sklearn.model_selection import train_test_split\nX_train,X_test,y_train,y_test=train_test_split(X,y,random_state=0,test_size=0.2)\nprint(X.shape, y.shape)\n\n# +\nlist_arg = [4, 5, 6]\n\n# args\ndef func(a, b, c):\n print(f\"a{a}, b{b}, c{c}\")\n\nfunc(1,2,3)\nfunc(*list_arg)\n\ndic_arg = {'a': 4,\n 'b' : 5,\n 'c' : 6}\n\n# kwargs\ndef func(a=0, b=1, c=2):\n print(f\"a{a}, b{b}, c{c}\")\n\nfunc(a=1,b=2)\nfunc(**dic_arg)\n# -\n\n# ## Applying regression model on training set\n\nfrom sklearn.ensemble import GradientBoostingRegressor\nparams = {'n_estimators': 130,\n 'max_depth': 5,\n 'min_samples_split': 2,\n 'learning_rate': 0.1,\n 'loss': 'ls'}\nregressor=GradientBoostingRegressor(**params)\nregressor.fit(X_train,y_train)\n# print(X.shape, y.shape)\n\n# ## Predicting the testing set\n\ny_pred=regressor.predict(X_test)\n\n# ## Calculating the accuracy of the model\n\n# +\nfrom sklearn.metrics import r2_score\n\nr2_score(y_test,y_pred)\n# -\n\n# # Explaining results\n\n# +\nimport matplotlib.pyplot as plt\nimport numpy as np\nfrom sklearn import datasets, ensemble\nfrom sklearn.inspection import permutation_importance\nfrom sklearn.metrics import mean_squared_error\nfrom sklearn.model_selection import train_test_split\n\nreg = regressor\n# -\n\ndataset.columns\n\nreg.feature_importances_\n\n# ## Plot training deviance\n\n# we visualize the results. To do that we will first compute the test set deviance and then plot it against boosting iterations.\n\n# +\ntest_score = np.zeros((params['n_estimators'],), dtype=np.float64)\nfor i, y_pred in enumerate(reg.staged_predict(X_test)):\n test_score[i] = reg.loss_(y_test, y_pred)\n\nfig = plt.figure(figsize=(6, 6))\nplt.subplot(1, 1, 1)\nplt.title('Deviance')\nplt.plot(np.arange(params['n_estimators']) + 1, reg.train_score_, 'b-',\n label='Training Set Deviance')\nplt.plot(np.arange(params['n_estimators']) + 1, test_score, 'r-',\n label='Test Set Deviance')\nplt.legend(loc='upper right')\nplt.xlabel('Boosting Iterations')\nplt.ylabel('Deviance')\nfig.tight_layout()\nplt.show()\n# -\n\n# ## Plot feature importance\n\n# impurity-based feature importances can be misleading for high cardinality features (many unique values)\n#\n# As an alternative, the permutation importances of reg can be computed on a held out test set. \n#\n# Permutation feature importance is a model inspection technique that can be used for any fitted estimator when the data is tabular (i.e. : in the form of an array)\n#\n#\n# The permutation feature importance is defined to be the decrease in a model score when a single feature value is randomly shuffled\n#\n# See Permutation feature importance for more details. https://scikit-learn.org/stable/modules/permutation_importance.html#permutation-importance\n\n# +\nfeature_names = dataset.drop(['smoker_no', 'smoker_yes'], axis=1).columns\n\nfeature_importance = reg.feature_importances_\nsorted_idx = np.argsort(feature_importance)\npos = np.arange(sorted_idx.shape[0]) + .5\nfig = plt.figure(figsize=(12, 6))\nplt.subplot(1, 2, 1)\nplt.barh(pos, feature_importance[sorted_idx], align='center')\nplt.yticks(pos, np.array(feature_names)[sorted_idx])\nplt.title('Feature Importance (MDI)')\n\nresult = permutation_importance(reg, X_test, y_test, n_repeats=10,\n random_state=42, n_jobs=2)\nsorted_idx = result.importances_mean.argsort()\nplt.subplot(1, 2, 2)\nplt.boxplot(result.importances[sorted_idx].T,\n vert=False, labels=np.array(feature_names)[sorted_idx])\nplt.title(\"Permutation Importance (test set)\")\nfig.tight_layout\n","repo_name":"ezalos/DataMining","sub_path":"9_Data_Exploration.ipynb","file_name":"9_Data_Exploration.ipynb","file_ext":"py","file_size_in_byte":8873,"program_lang":"python","lang":"en","doc_type":"code","stars":0,"dataset":"github-jupyter-script","pt":"41"}
+{"seq_id":"20661998520","text":"# import os to interact with the operating systen\nimport os\nimport numpy as np\nimport pandas as pd\nimport matplotlib.pyplot as plt\n\n\nos.getcwd()\n\nos.chdir('C:\\\\Users\\\\user\\\\Desktop\\\\data science\\\\data\\\\2021')\n\npath_2021 = os.getcwd()\n\n# +\n# import of the excel files for analysis\n\ngrade8 = pd.read_excel(\"08.xls\")\ngrade9 = pd.read_excel(\"09.xls\")\ngrade10 = pd.read_excel(\"10.xls\")\ngrade11 = pd.read_excel(\"11.xls\")\n\n# +\n# preparing grade 8 file\n# removing unnecesary rows and columns\n\nmust_go_col8= grade8.loc[:, 'Unnamed: 31':'Unnamed: 119']\n# g8 = grade8.iloc[6:-10,:]\n# grade8.head()\n\n# +\n# preparing grade 9 file\n# removing unnecesary rows and columns\n\nmust_go_col9= grade9.loc[:, 'Unnamed: 31':'Unnamed: 119']\n# must_go_row = grade9.iloc[6:-10,:]\n# must_go_row.head()\n\n# +\n# preparing grade 10 file\n# getting rid of unnecessary columns\n\n\nmust_go_col10= grade10.loc[:, 'Unnamed: 39':'Unnamed: 119']\n# must_go_row10 = grade10.iloc[6:-10,:]\n# must_go_row10.head()\n\n# +\n# preparing grade 11 file\n# getting rid of unnecessary columns\n\n\nmust_go_col11= grade11.loc[:, 'Unnamed: 39':'Unnamed: 121']\n# must_go_row11.head()\n\n# +\n# ucols = ['must_go_col8', 'must_go_col9', 'must_go_col10', 'must_go_col11']\n\n# +\n\ngrade11.columns\n# g9.columns\n# -\n\ngrade10.drop(must_go_col10, axis='columns', inplace=True)\ngrade11.drop(must_go_col11, axis='columns', inplace=True)\n\ngrade10.columns\n\ngrade11.columns\n\n# +\n# unnecessary columns for GET\n\nunwanted_subjects = [' Level', ' Level', ' Level.2',' Level.4',' Level.5',' Level.6', ' Level.7',' Level.8', ' Level.1',' Home Language (2)',' Level.1',' 1st Additional\\n Language(2)', ' Level.2', ' 1st Additional\\n Language(3)', ' Level.3']\n\n# +\n# getting rid of unnecessary columns\n\ngrade8.drop(must_go_col8, axis='columns', inplace=True)\ngrade9.drop(must_go_col9, axis='columns', inplace=True)\n\ngrade8.drop(unwanted_subjects, axis='columns', inplace=True)\ngrade9.drop(unwanted_subjects, axis='columns', inplace=True)\n# -\n\ngrade9.columns\n\ngrade8.columns\n\n# +\n# grade8.head()\n\n# +\n# grade8.columns\n# -\n\n# # Grade 8 Analysis\n\ng8 = grade8[[\"Learner\", \"Avg\", \"code\"]]\ng8\n\n# # Pass and fail\n\n# +\n# counting learners who passed and who failed\n\npassed = 0\nfailed = 0\n\nfor i in g8.code:\n if i == \"P\":\n passed += 1\n else:\n failed +=1\n \nprint(\"Passed: \", passed)\nprint(\"Failed :\", failed)\n# -\n\n# listing learners who failed\n\ng8fail = g8[g8.code == \"NP\"]\ng8fail\n\n# # Failed Maths, Setswana and/or English\n\n# +\n# failed mathematics\n\nmaths = grade8[grade8[\"Mathematics\\n(Gr 08)\"] < 40 ]\n\nmat = maths[[\"Learner\", \"Mathematics\\n(Gr 08)\"]]\n\n\n# failed setswana\n\nsetswana = grade8[grade8[' Home Language'] < 50 ]\n\nsets = setswana[[\"Learner\", \" Home Language\"]]\n\n\n# failed English\n\nenglish = grade8[grade8[' First Additional Language'] < 40 ]\n\neng = english[[\"Learner\", \" First Additional Language\"]]\n\n\nprint(\"______________________________________________________\")\nprint(\"Failed mathematics\")\nprint(\"______________________________________________________\")\nprint(mat)\n\nprint(\"______________________________________________________\")\nprint(\"\\n Failed Setswana\")\nprint(\"______________________________________________________\")\nprint(sets)\n\nprint(\"______________________________________________________\")\nprint(\"\\n Failed English FAL\")\nprint(\"______________________________________________________\")\nprint(eng)\nprint(\"______________________________________________________\")\n\n\n\n# +\n# add 5% to setswana\n\n# sets = sets[\" Home Language\"] + 5\n# sets\n# -\n\n# # working with subjects\n\nsubjects =[' Home Language',' First Additional Language', 'Creative Arts\\n(Gr 08)', 'Economic Management Sciences\\n(Gr 08)', 'Life Orientation\\n(Gr 08)', 'Mathematics\\n(Gr 08)','Natural Sciences\\n(Gr 08)', 'Social Sciences\\n(Gr 08)', 'Technology\\n(Gr 08)' ]\n\n# +\n# abbreviated subjects\n\nabv_sub = ['HL', 'FAL', 'CA', 'EMS', 'LO', 'MAT','NS', 'SS', 'TECH']\n# -\n\navg = []\nfor i in subjects:\n i = grade8[i].mean()\n avg.append(i)\n\navg = []\nfor i in subjects:\n i = grade8[i].mean()\n avg.append(i)\n\n# +\nfig = plt.figure(figsize=(16, 8))\n\nplt.bar(abv_sub, avg, color = \"brown\")\nplt.xlabel(\"Subjects\")\nplt.ylabel(\"Average\")\nplt.title(\"Learner performance in different Subjects\")\n# -\n\n# # Awards\n\n# Top 10\n\ntop_10 = g8.sort_values(by=(\"Avg\"), ascending=False)\ntop_10.head(10)\n\n# best per subject\n\n# +\ncols = grade8.columns\n\nsub = 6\n\nfor col in cols:\n if sub < 15:\n sub = sub + 1\n subject = cols[sub]\n \n top = grade8[['Learner', subject]]\n best = top.sort_values(by=(subject), ascending=False)\n print(best.head(1))\n print(\"------------------------------------------------------------------\")\n print()\n# -\n\n# # Grade 9 Analysis\n\ng9 = grade9[[\"Learner\", \"Avg\", \"code\"]]\ng9.head()\n\n# # number passed and failed\n\n# +\n# counting learners who passed and who failed\n\npassed = 0\nfailed = 0\n\nfor i in g9.code:\n if i == \"P\":\n passed += 1\n else:\n failed +=1\n \nprint(\"Passed: \", passed)\nprint(\"Failed :\", failed)\n# -\n\n# Learners who failed\n\ng9fail = g9[g9.code == \"NP\"]\ng9fail\n\n# # Failed Maths, Setswana and/or English\n\n# +\n# failed mathematics\n\nmaths9 = grade9[grade9[\"Mathematics\\n(Gr 09)\"] < 40 ]\n\nmat9 = maths9[[\"Learner\", \"Mathematics\\n(Gr 09)\"]]\n\n\n# failed setswana\n\nsetswana9 = grade9[grade9[' Home Language'] < 50 ]\n\nset9 = setswana9[[\"Learner\", \" Home Language\"]]\nset9\n\n# failed English\n\nenglish9 = grade9[grade9[' First Additional Language'] < 40 ]\n\neng9 = english9[[\"Learner\", \" First Additional Language\"]]\neng9\n\n\n\nprint(\"____________________________________________________________\")\nprint(\"Failed mathematics\")\nprint(\"------------------------------------------------------------\")\nprint(mat9)\n\nprint(\"____________________________________________________________\")\nprint(\"\\n Failed Setswana\")\nprint(\"------------------------------------------------------------\")\nprint(set9)\n\nprint(\"____________________________________________________________\")\nprint(\"\\n Failed English FAL\")\nprint(\"------------------------------------------------------------\")\nprint(eng9)\nprint(\"____________________________________________________________\")\n\n# -\n\n# g8fail = g8[g8.code == \"NP\"]\n# g8fail\ngrade9.columns\n\n\n# # working with subjects\n\nsubjects_9 =[' Home Language',' First Additional Language', 'Creative Arts\\n(Gr 09)', 'Economic Management Sciences\\n(Gr 09)', 'Life Orientation\\n(Gr 09)', 'Mathematics\\n(Gr 09)','Natural Sciences\\n(Gr 09)', 'Social Sciences\\n(Gr 09)', 'Technology\\n(Gr 09)' ]\n\navg_9 = []\nfor i in subjects_9:\n i = grade9[i].mean()\n avg_9.append(i)\n\n# +\nfig = plt.figure(figsize=(16, 8))\n\nplt.bar(abv_sub, avg_9)\nplt.xlabel(\"Subjects\")\nplt.ylabel(\"Average\")\nplt.title(\"Learner performance in different Subjects\")\n# -\n\n# # Awards\n\n# Top 10\n\ntop_10_9 = g9.sort_values(by=(\"Avg\"), ascending=False)\ntop_10_9.head(10)\n\n# best per subject\n\n# +\ncols = grade9.columns\n\nsub = 6\n\nfor col in cols:\n if sub < 15:\n sub = sub + 1\n subject = cols[sub]\n \n top = grade9[['Learner', subject]]\n best = top.sort_values(by=(subject), ascending=False)\n print(best.head(1))\n print(\"------------------------------------------------------------------\")\n print()\n","repo_name":"tswelopelemojewa/data-analysis","sub_path":"senior phase analysis.ipynb","file_name":"senior phase analysis.ipynb","file_ext":"py","file_size_in_byte":7214,"program_lang":"python","lang":"en","doc_type":"code","stars":0,"dataset":"github-jupyter-script","pt":"40"}
+{"seq_id":"42282981208","text":"# + id=\"9vYKYluW6DE7\" colab_type=\"code\" colab={}\nimport numpy as np\nimport pandas as pd\nimport matplotlib.pyplot as plt\nimport seaborn as sn\n\nimport librosa, librosa.display\nimport IPython.display as ipd\nfrom scipy.io import wavfile \n\nfrom sklearn.linear_model import LogisticRegression\nfrom sklearn.metrics import classification_report, confusion_matrix\n\nfrom google.colab import drive\nimport pickle\n\n# + id=\"mtaLEKw96PU3\" colab_type=\"code\" outputId=\"d9b60ee9-d242-4dfd-ea54-c890c8950488\" executionInfo={\"status\": \"ok\", \"timestamp\": 1585209071619, \"user_tz\": -180, \"elapsed\": 29415, \"user\": {\"displayName\": \"Elena Timofeeva\", \"photoUrl\": \"\", \"userId\": \"07958490506212570356\"}} colab={\"base_uri\": \"https://localhost:8080/\", \"height\": 122}\ndrive.mount('/content/drive')\n\n\n# + id=\"DDf1CCrx6SJa\" colab_type=\"code\" colab={}\ndef load_pkl_data():\n X_train = pickle.load(open('drive/My Drive/Audio analytics/pickle/X_train.pkl', 'rb'))\n y_train = pickle.load(open('drive/My Drive/Audio analytics/pickle/y_train.pkl', 'rb'))\n X_test = pickle.load(open('drive/My Drive/Audio analytics/pickle/X_test.pkl', 'rb'))\n y_test = pickle.load(open('drive/My Drive/Audio analytics/pickle/y_test.pkl', 'rb'))\n\n return X_train, y_train, X_test, y_test\n\n\n# + id=\"wuD4OpHP-7dO\" colab_type=\"code\" colab={}\n# Приведение к одномерному массиву\ndef changing_dim(X):\n\n for i in range(len(X)):\n X[i] = np.ravel(X[i])\n\n # Приведение массивов к одной длине (заполение нулями)\n max_len = max([len(i) for i in X])\n for i in range(len(X)):\n X[i] = np.resize(X[i], max_len)\n\n return X\n\n\n# + id=\"QqWmp39X7UTA\" colab_type=\"code\" colab={}\nX_train, y_train, X_test, y_test = load_pkl_data()\nclass_names = list(set(y_test))\n\n# + id=\"bPXOYWVJ7ZMP\" colab_type=\"code\" colab={}\nX_train = changing_dim(X_train)\nX_test = changing_dim(X_test)\n\n# + id=\"yqe0-eakAE-A\" colab_type=\"code\" outputId=\"097d238b-6c0f-4461-93a7-84da987c59ec\" executionInfo={\"status\": \"ok\", \"timestamp\": 1585216601699, \"user_tz\": -180, \"elapsed\": 603130, \"user\": {\"displayName\": \"Elena Timofeeva\", \"photoUrl\": \"\", \"userId\": \"07958490506212570356\"}} colab={\"base_uri\": \"https://localhost:8080/\", \"height\": 51}\n# Классификация с помощью Логистической регрессии\nlog_reg = LogisticRegression(solver='saga')\nlog_reg.fit(X_train, y_train)\ny_pred = log_reg.predict(X_test)\n\n# + id=\"zZdgTWBJHtMt\" colab_type=\"code\" outputId=\"4f57a43b-e8a3-42de-fd42-40707a40ab41\" executionInfo={\"status\": \"ok\", \"timestamp\": 1585217045884, \"user_tz\": -180, \"elapsed\": 630, \"user\": {\"displayName\": \"Elena Timofeeva\", \"photoUrl\": \"\", \"userId\": \"07958490506212570356\"}} colab={\"base_uri\": \"https://localhost:8080/\", \"height\": 833}\n# Оценки для каждого класса\nprint(classification_report(y_test, y_pred, labels=class_names))\n\n# + id=\"ZpwdS7zPEJfo\" colab_type=\"code\" outputId=\"751316aa-0f48-43c3-c74c-8e52643ac52b\" executionInfo={\"status\": \"ok\", \"timestamp\": 1585216804036, \"user_tz\": -180, \"elapsed\": 6526, \"user\": {\"displayName\": \"Elena Timofeeva\", \"photoUrl\": \"\", \"userId\": \"07958490506212570356\"}} colab={\"base_uri\": \"https://localhost:8080/\", \"height\": 1000}\n# Построение Confusion matrix\nconf_matrix = confusion_matrix(y_test, y_pred, labels=class_names)\ndf_cm = pd.DataFrame(conf_matrix, index=class_names, columns=class_names)\n\nplt.figure(figsize=(15,15))\nsn.set(font_scale=1.4)\nsn.heatmap(df_cm, annot=True, annot_kws={\"size\": 15}, cmap='Blues')\n\n# + id=\"XEWHht1KFVW6\" colab_type=\"code\" colab={}\n\n","repo_name":"LenaTim/root","sub_path":"LogisticRegression.ipynb","file_name":"LogisticRegression.ipynb","file_ext":"py","file_size_in_byte":3563,"program_lang":"python","lang":"en","doc_type":"code","stars":0,"dataset":"github-jupyter-script","pt":"40"}
+{"seq_id":"24946603619","text":"# +\n# %matplotlib inline\nimport matplotlib.pyplot as plt\nimport numpy as np\nimport math\n\n#We now obtain the ndarray object of angles between 0 and 2π using the arange() function from the NumPy library.\nx = np.arange(0, math.pi*2, 1)\n\n#The ndarray object serves as values on x axis of the graph. \n#The corresponding sine values of angles in x to be displayed on y axis are obtained by the following statement −\ny = np.sin(x)\n\n#The values from two arrays are plotted using the plot() function.\nplt.plot(x,y)\n\n#You can set the plot title, and labels for x and y axes.\nplt.xlabel(\"angle\")\nplt.ylabel(\"sine\")\nplt.title('sine wave')\n\n#The Plot viewer window is invoked by the show() function −\nplt.show()\n# -\n\n\n\n\n","repo_name":"akashreveluv/Python-Tutorials","sub_path":"python_matplotlib/1-simple plot.ipynb","file_name":"1-simple plot.ipynb","file_ext":"py","file_size_in_byte":712,"program_lang":"python","lang":"en","doc_type":"code","stars":0,"dataset":"github-jupyter-script","pt":"40"}
+{"seq_id":"9107630760","text":"# ## Shared Bikes Demand Prediction - BoomBikes\n#\n# #### Problem Statement:\n#\n# A US bike-sharing provider `BoomBikes` has a daily dataset on the rental bikes based on various environmental and seasonal settings. It wishes to use this data to understand the factors affecting the demand for these shared bikes in the American market and come up with a mindful business plan to be able to accelerate its revenue as soon as the ongoing lockdown due to corona pandemic comes to an end.\n#\n# **Essentially, the company wants to know**:\n#\n#\n# - Which variables are significant in predicting the demand for shared bikes.\n#\n#\n# - How well those variables describe the bike demands\n#\n\n# **Overall Step to Build a ML model**\n#\n# \n\n# This solution is divided into the following main sections: \n# - Data understanding/exploration\n# - Data Visualisation \n# - Data preparation/cleaning\n# - Model building and evaluation\n# - Residual Analysis (verifying the assumptions of Linear Model)\n#\n\n# ### 1. Data Understanding/Exploration\n#\n# Let's first import the required libraries and have a look at the dataset and understand the size, attribute names etc.\n\nimport numpy as np\nimport pandas as pd\nimport matplotlib.pyplot as plt\nimport seaborn as sns\nfrom sklearn import linear_model\nfrom sklearn.linear_model import LinearRegression\n\nimport warnings\nwarnings.filterwarnings('ignore')\n\n# Reading the dataset\nbike = pd.read_csv(\"day.csv\")\n\n# Let's take a look at the first few rows\nbike.head(3).append(bike.tail(3))\n\n# Let's look at the number of rows and columns in the dataset\nbike.shape\n\n# Understanding the feature names in the dataset\nprint(bike.columns)\nbike.dtypes.value_counts()\n\n# Summary of the dataset: 730 rows, 16 columns, no null values\nprint(bike.info())\n\n# Getting insights of the features\nbike.describe()\n\n# #### Understanding the Data Dictionary and parts of Data Preparation\n#\n# The data dictionary contains the meaning of various attributes; some of which are explored and manipulated here:\n\n# +\n# Assigning string values to different seasons instead of numeric values. These numeric values may misindicate some order to it.\nprint(bike['season'].value_counts())\n# 1=spring\nbike.loc[(bike['season'] == 1) , 'season'] = 'spring'\n\n# 2=summer\nbike.loc[(bike['season'] == 2) , 'season'] = 'summer'\n\n# 3=fall\nbike.loc[(bike['season'] == 3) , 'season'] = 'fall'\n\n# 4=winter\nbike.loc[(bike['season'] == 4) , 'season'] = 'winter'\n# -\n\n# Checking whether the conversion is done properly or not and getting data count on the basis of season\nbike['season'].astype('category').value_counts()\n\n# year (0: 2018, 1:2019)\nbike['yr'].astype('category').value_counts()\n\n# +\n# Assigning string values to different months instead of numeric values which may misindicate some order to it.\n# A function has been created to map the actual numbers to categorical levels.\n# Check the data before change\nprint(bike['mnth'].astype('category').value_counts())\ndef object_map(x):\n return x.map({1: 'Jan', 2: 'Feb', 3: 'Mar', 4: 'Apr', 5: 'May', 6: 'Jun', 7: 'Jul',8: 'Aug',9: 'Sept',10: 'Oct',11: 'Nov',12: 'Dec'})\n\n# Applying the function to the two columns\nbike[['mnth']] = bike[['mnth']].apply(object_map)\n# -\n\n# Checking whether the conversion is done properly or not and getting data count on the basis of month\nbike['mnth'].astype('category').value_counts(ascending = True)\n\n# whether day is a holiday or not (0: No, 1: Yes)\nbike['holiday'].astype('category').value_counts()\n\n\n# +\n# Assigning string values to weekdays instead of numeric values. These values may misindicate some order to it.\n# A function has been created to map the actual numbers to categorical levels.\ndef str_map(x):\n return x.map({1: 'Wed', 2: 'Thur', 3: 'Fri', 4: 'Sat', 5: 'Sun', 6: 'Mon', 0: 'Tue'})\n\n# Applying the function to the two columns\nbike[['weekday']] = bike[['weekday']].apply(str_map)\n# -\n\n# Checking whether the conversion is done properly or not and getting data count on the basis of weekdays\nbike['weekday'].astype('category').value_counts()\n\n# if a day is neither weekend nor a holiday it takes the value 1, otherwise 0\nbike['workingday'].astype('category').value_counts()\n\n# +\n# Optional: Replacing long weathersit names into string values for better readability and understanding\n\n# 1-Clear, Few clouds, Partly cloudy, Partly cloudy\nbike.loc[(bike['weathersit'] == 1) , 'weathersit'] = 'A'\n\n# 2-Mist + Cloudy, Mist + Broken clouds, Mist + Few clouds, Mist\nbike.loc[(bike['weathersit'] == 2) , 'weathersit'] = 'B'\n\n# 3-Light Snow, Light Rain + Thunderstorm + Scattered clouds, Light Rain + Scattered clouds\nbike.loc[(bike['weathersit'] == 3) , 'weathersit'] = 'C'\n\n# 4-Heavy Rain + Ice Pallets + Thunderstorm + Mist, Snow + Fog\nbike.loc[(bike['weathersit'] == 4) , 'weathersit'] = 'D'\n# -\n\n# Extracting the type of weather situations present in the data\nprint(bike['weathersit'].unique())\nprint(bike['weathersit'].astype('category').value_counts(ascending = True))\n\n# Converting date to datetime format\nprint(\"As is:\",bike['dteday'].dtypes)\nbike['dteday']=bike['dteday'].astype('datetime64')\nprint(\"Now:\", bike['dteday'].dtypes)\n\n# ### Conclusion: 1. Data Understanding/Exploration\n#\n# In this section, we have analysed the given dataset w.r.to it's structure. In the process, we have changed the data types of few columns and also changed some of the values based on Data Dictionary.\n\n# ### 2. Data Visualisation\n#\n# Let's now spend some time doing what is arguably the most important step - **Data Content Analysis**.\n# - Understanding the distribution of various numeric variables \n# - If there is some obvious multicollinearity going on, this is the first place to catch it\n# - Here's where you'll also identify if some predictors directly have a strong association with the outcome variable\n#\n# We'll visualise our data using `matplotlib` and `seaborn`.\n\n# temp: temperature in Celsius\nsns.set_style(\"darkgrid\")\nsns.distplot(bike['temp'],bins = 20, color = 'b').set(title='Temperature in Celsius')\nplt.show()\n\n# **Q1. What is Density here?**\n# - **Q2. What will happen if I change the bins to 5 or 55? Will the graph shows a Normal Distribution?**\n#\n\n# feeling temperature\nsns.distplot(bike['atemp']).set(title='Feeling Temperature')\nplt.show()\n\n# humidity\nsns.distplot(bike['hum'], color = 'g').set(title='Humidity')\nplt.show()\n\n# wind speed\nsns.distplot(bike['windspeed'], color = 'r').set(title='Wind Speed')\nplt.show()\n\n# #### Let's analyse our dependent variables: cnt\n\n# Target variable: count of total rental bikes including both casual and registered\nsns.distplot(bike['cnt'], bins = 10).set(title='Count of Rented Bikes')\nplt.show()\n\n# 0: 2018, 1: 2019\nsns.barplot(x='yr', y='cnt', data=bike)\nplt.show()\n\nsns.barplot(x='weekday', y='cnt', data=bike)\nplt.show()\n\n# Holiday - 0: No, 1: Yes\nsns.barplot(x='holiday', y='cnt', data=bike)\nplt.show()\n\n# +\n# weathersit\n# 'A'- Clear, Few clouds, Partly cloudy, Partly cloudy\n# 'B'- Mist + Cloudy, Mist + Broken clouds, Mist + Few clouds, Mist\n# 'C'- Light Snow, Light Rain + Thunderstorm + Scattered clouds, Light Rain + Scattered clouds\n# 'D'- Heavy Rain + Ice Pallets + Thunderstorm + Mist, Snow + Fog\n\nsns.barplot(x='weathersit', y='cnt', data=bike)\nplt.show()\n# -\n\nbike.info()\n\n# All categorical variables in the dataset\nbike_categorical = bike.select_dtypes(exclude=['float64','datetime64','int64'])\nprint(\"Only categorical varibales:\", bike_categorical.columns)\n\nbike_categorical.head(3).append(bike_categorical.tail(3))\n\nbike.dtypes.value_counts()\n\n# #### Visualising Categorical Variables\n#\n# As you might have noticed, there are a few categorical variables as well. Let's make a boxplot for some of these variables.\n\nplt.figure(figsize=(20, 20)) \nplt.subplot(3,3,1)\nsns.boxplot(x = 'season', y = 'cnt', data = bike)\nplt.subplot(3,3,2)\nsns.boxplot(x = 'holiday', y = 'cnt', data = bike)\nplt.subplot(3,3,3)\nsns.boxplot(x = 'workingday', y = 'cnt', data = bike)\nplt.subplot(3,3,4)\nsns.boxplot(x = 'weathersit', y = 'cnt', data = bike)\nplt.subplot(3,3,5)\nsns.boxplot(x = 'mnth', y = 'cnt', data = bike)\nplt.subplot(3,3,6)\nsns.boxplot(x = 'weekday', y = 'cnt', data = bike)\nplt.subplot(3,3,7)\nsns.boxplot(x = 'yr', y = 'cnt', data = bike)\nplt.show()\n\n# #### Visualising Numeric Variables\n#\n# Let's make a pairplot of all the numeric variables\n\n# +\n# Converting \"casual\",\"registered\" and \"cnt\" Integer/discreate variables to float. \n# This step is performed to seperate out categorical variables like 'yr','holiday','workingday' which have binary values in them\nInt_Var_List = [\"casual\",\"registered\",\"cnt\"]\n\nfor var in Int_Var_List:\n bike[var] = bike[var].astype(\"float\")\nbike[Int_Var_List].head(3)\n# -\n\n# All numeric variables in the dataset\nbike_numeric = bike.select_dtypes(include=['float64'])\nbike_numeric.head()\n\n# +\ncolors = iter(['xkcd:red purple', 'xkcd:pale teal', 'xkcd:warm purple',\n 'xkcd:light forest green', 'xkcd:blue with a hint of purple',\n 'xkcd:light peach', 'xkcd:dusky purple', 'xkcd:pale mauve',\n 'xkcd:bright sky blue', 'xkcd:baby poop green', 'xkcd:brownish',\n 'xkcd:moss green', 'xkcd:deep blue', 'xkcd:melon',\n 'xkcd:faded green', 'xkcd:cyan', 'xkcd:brown green',\n 'xkcd:purple blue', 'xkcd:baby shit green','xkcd:red purple', 'xkcd:pale teal', 'xkcd:warm purple',\n 'xkcd:light forest green', 'xkcd:blue with a hint of purple',\n 'xkcd:light peach', 'xkcd:dusky purple', 'xkcd:pale mauve',\n 'xkcd:bright sky blue', 'xkcd:baby poop green', 'xkcd:brownish',\n 'xkcd:moss green', 'xkcd:deep blue', 'xkcd:melon',\n 'xkcd:faded green', 'xkcd:cyan', 'xkcd:brown green',\n 'xkcd:purple blue', 'xkcd:baby shit green','xkcd:red purple', 'xkcd:pale teal', 'xkcd:warm purple',\n 'xkcd:light forest green', 'xkcd:blue with a hint of purple',\n 'xkcd:light peach', 'xkcd:dusky purple', 'xkcd:pale mauve',\n 'xkcd:bright sky blue', 'xkcd:baby poop green', 'xkcd:brownish',\n 'xkcd:moss green', 'xkcd:deep blue', 'xkcd:melon',\n 'xkcd:faded green', 'xkcd:cyan', 'xkcd:brown green',\n 'xkcd:purple blue', 'xkcd:baby shit green','xkcd:red purple', 'xkcd:pale teal', 'xkcd:warm purple',\n 'xkcd:light forest green', 'xkcd:blue with a hint of purple',\n 'xkcd:light peach', 'xkcd:dusky purple', 'xkcd:pale mauve',\n 'xkcd:bright sky blue', 'xkcd:baby poop green', 'xkcd:brownish',\n 'xkcd:moss green', 'xkcd:deep blue', 'xkcd:melon',\n 'xkcd:faded green', 'xkcd:cyan', 'xkcd:brown green',\n 'xkcd:purple blue', 'xkcd:baby shit green', ])\n\ndef modified_scatter(x,y, **kwargs):\n kwargs['color'] = next(colors)\n plt.scatter(x,y, **kwargs)\n \n\n\n# +\n# Pairwise scatter plot\ng = sns.pairplot(bike_numeric)\n\ng.map_offdiag(modified_scatter)\nplt.show()\n# -\n\n# We can better plot correlation matrix between variables to know the exact values of correlation between them. Also, a heatmap is pretty useful to visualise multiple correlations in one plot.\n\n# Correlation matrix\ncor = bike_numeric.corr()\ncor\n\n# Let's plot the correlations on a heatmap for better visualisation\n\n# heatmap\nmask = np.array(cor)\nmask[np.tril_indices_from(mask)] = False\nfig,ax= plt.subplots()\nfig.set_size_inches(12,7)\nsns.heatmap(cor, mask=mask, cmap = 'PiYG', annot=True)\n\n# The heatmap shows some useful insights:\n#\n# Correlation of Count('cnt') with independent variables:\n# - Count('cnt') is highly (positively) correlated with 'casual' and 'registered' and further it is high with 'atemp'. We can clearly understand the high positive correlation of count with 'registered' and 'casual' as both of them together add up to represent count.\n#\n# - Count is negatively correlated to 'windspeed' (-0.24 approximately). This gives us an impression that the shared bikes demand will be somewhat less on windy days as compared to normal days.\n# - correlation with \"humidity\" and \"cnt\" is almost negligible (- 0.09)\n#\n# Correlation among independent variables:\n# - Some of the independent variables are highly correlated (look at the top-left part of matrix): atemp and temp are highly (positively) correlated. The correlation between the two is almost equal to 1 - Suggesting that these two variables almost always give similar information. \n#\n#\n# Thus, while building the model, we'll have to pay attention to **multicollinearity**.\n\n#removing atemp as it is highly correlated with temp\nbike.drop('atemp',axis=1,inplace=True) \n\nprint(bike.shape)\nprint(bike.dtypes.value_counts())\n\n# ### Conclusion: 2. Data Visualization\n#\n# In this section, we have looked at the actual data content.We plotted few distribution plots and a correleation Matrix and heatmap for numerical varibales. We also removed one of the features \"atemp\" as it was highly correlated with the another varibales \"temp\". \n\n# ## 3. Data Preparation \n#\n#\n# #### Data Preparation\n#\n# Let's now prepare the data and build the model.\n# Note that we had not included 'yr', 'mnth', 'holiday', 'weekday' and 'workingday' as object variables in the initial data exploration steps so as to avoid too many dummy variables creation. They have binary values: 0s and 1s in them which have specific meanings associated with them.\n\n# Subset all categorical variables\nbike_categorical=bike.select_dtypes(include=['object'])\nbike_categorical.head()\n\n# #### Dummy Variables\n# The variable `season`,`mnth`,`weekday` and `weathersit` have different levels. We need to convert these levels into integers. \n#\n# For this, we will use something called `dummy variables`.\n\n# Convert into dummies\nbike_dummies = pd.get_dummies(bike_categorical, drop_first=True)\nbike_dummies.head()\n\nbike_dummies.columns\n\n# This code block display how the data has been encoded in dummy variables.\n# For weathersit A, where no variables has been created, the value under weathersit_B and weathersit_C will be both 0.\npd.set_option('display.max_columns', None)\nprint(bike.iloc[25:27])\nprint(\"----------------------------------------\")\nprint(bike_dummies.iloc[25:27])\n\n# #### Q. Why do we use drop_first=True during dummy variable creation? Does it really matters?\n# - Yes, it is important to use **drop_first = True** while creating the dummy variables. It helps in reducing the extra column created during dummy variable creation. Hence it reduces the correlations created among dummy variables. Hence if we have categorical variable with N distinc values, then we need only N-1 columns to represent the dummy variables.\n#\n\n# Drop categorical variable columns\nbike = bike.drop(list(bike_categorical.columns), axis=1)\n\n# Concatenate dummy variables with the original dataframe\nbike = pd.concat([bike, bike_dummies], axis=1)\n\n# Let's check the first few rows\nbike.head()\n\n# +\n# Check Column 'instant'[Record Index] and 'dteday'[Date]\n\nprint(len(bike.instant.unique()))\nprint(len(bike.dteday.unique()))\n# -\n\n# Drop the 'instant' and 'dteday' column as they of not any use to us for the analysis\nbike=bike.drop(['instant','dteday'], axis = 1, inplace = False)\nbike.head(3).append(bike.tail(3))\n\nprint(bike.head(3))\n\n# ### Conclusion: 3. Data Preparation\n#\n# The main task done in this section are:\n# - create the respective dummy variables\n# - delete the irrelevant data\n\n# ## 4. Model Building and Evaluation\n#\n# Let's start building the model. The first step to model building is the usual test-train split. So let's perform that\n\n# Before splitting, make a copy of the cleaned data frame\nbike_c = bike.copy()\nprint(bike.shape)\nprint(bike_c.shape)\n\n# +\n# Split the dataframe into train and test sets\nfrom sklearn.model_selection import train_test_split\n\ndf_train, df_test = train_test_split(bike, train_size=0.7, test_size=0.3, random_state=100)\n# -\n\nprint(df_train.shape)\ndf_train.head(10)\n\n# ### Scaling\n#\n# - for better interpretability\n# - only the numeric columns and not the dummy variables\n# - only on the train dataset as you don't want it to learn anything from the test data.\n\n# Let's scale all these columns using **MinMaxScaler**. You can use any other scaling method like **Standardization**\n\nfrom sklearn.preprocessing import MinMaxScaler \n# from sklearn.preprocessing import StandardScaler - in case you want to use Standardization method\n\nscaler = MinMaxScaler()\n\nbike_numeric.columns # created in previous section\n\n# +\n# Apply scaler() to all the columns except the 'yes-no' and 'dummy' variables\nvar = ['temp', 'hum', 'windspeed','casual','registered','cnt']\n\ndf_train[var] = scaler.fit_transform(df_train[var])\n\n# +\ndf_train[var].describe()\n\n# Please note that min is 0 and max. is 1 now\n# -\n\n# As expected, the variables have been appropriately scaled.\n\ndf_train.describe()\n\ndf_train.corr()\n\n# Let's check the correlation coefficients to see which variables are highly correlated\nplt.figure(figsize = (30, 30))\nsns.heatmap(df_train.corr(), annot = True, cmap=\"YlGnBu\")\nplt.show()\n\n# As we can noticed, `temp` seems to the correlated to `cnt` the most, after 'casual' and 'registered'. Let's see a pairplot for `temp` vs `cnt`. The above heatmap is not much readable, so alternatively, we can create a label for the correlation coefficients to highlight the certain values.\n\ndf_corr = df_train.corr(method=\"pearson\")\n\n# +\n# Create labels for the correlation matrix\nlabels = np.where(np.abs(df_corr)>0.75, \"H\",\n np.where(np.abs(df_corr)>0.5, \"M\",\n np.where(np.abs(df_corr)>0.25, \"S\", \"\")))\n\n# Plot correlation matrix\nplt.figure(figsize=(15, 15))\nsns.heatmap(df_corr, mask=np.eye(len(df_corr)), square=True,\n center=0, annot=labels, fmt='', linewidths=.5,\n cmap=\"PiYG\", cbar_kws={\"shrink\": 0.5});\n\n# +\nplt.figure(figsize=(20, 20)) \n\nplt.subplot(6,6,1)\nplt.scatter(df_train.temp, df_train.cnt)\n\nplt.subplot(6,6,2)\nplt.scatter(df_train.casual, df_train.cnt)\n\nplt.subplot(6,6,3)\nplt.scatter(df_train.registered, df_train.cnt)\nplt.show()\n# -\n\n# #### Dividing into X and Y sets for the model building\n\n# Dropping 'casual' and 'registered' as together they add up to cnt\ny_train = df_train.pop('cnt')\nX_train = df_train.drop([\"casual\",\"registered\"],axis=1) \n\nprint(X_train.shape)\nX_train.head()\n\ny_train.shape\n\n# ### Building the first model with all the features\n#\n# Let's now build our first model with all the features.\n\n# +\nimport statsmodels.api as sm\nX_train_lm = sm.add_constant(X_train)\nlr_model1 = sm.OLS(y_train, X_train_lm).fit()\n\nlr_model1.params\n\n# +\n# Creating a model using sklearn Linear Regression\nlm = LinearRegression()\n\n# Fit a line\nlm.fit(X_train, y_train)\n# -\n\n# Print the coefficients and intercept\nprint(lm.coef_)\nprint(lm.intercept_)\n\n# getting the model summary from statsmodel\nlr_model1.summary()\n\n# **How to Read the Output**\n\n# Please refer to these blogs:\n# - https://medium.com/swlh/interpreting-linear-regression-through-statsmodels-summary-4796d359035a\n# - https://www.adrian.idv.hk/2021-07-16-statsmodels/\n\n# This model has an Adjusted R-squared value of **84.5%** which seems pretty good. But let's see if we can reduce the number of features and exclude those which are not much relevant in explaining the target variable. \n\n# #### Model Building Using RFE\n#\n# Now, you have close to 28 features. It is obviously not recommended to manually eliminate these features. So let's now build a model using recursive feature elimination to select features. We'll first start off with an arbitrary number of features (15 seems to be a good number to begin with), and then use the `statsmodels` library to build models using the shortlisted features (this is also because `SKLearn` doesn't have `Adjusted R-squared` that `statsmodels` has).\n\n# +\n# Import RFE\nfrom sklearn.feature_selection import RFE\n\n# RFE with 15 features\nlm = LinearRegression()\nrfe1 = RFE(lm, 15)\n\n# Fit with 15 features\nrfe1.fit(X_train, y_train)\n\n# Print the boolean results\nprint(rfe1.support_) \nprint(rfe1.ranking_) \n# -\n\n# #### Model Building and Evaluation \n#\n# Let's now check the summary of this model using `statsmodels`.\n\n# +\n# Import statsmodels\nimport statsmodels.api as sm \n\n# Subset the features selected by rfe1\ncol1 = X_train.columns[rfe1.support_]\n\n# Subsetting training data for 15 selected columns\nX_train_rfe1 = X_train[col1]\n\n# Add a constant to the model\nX_train_rfe1 = sm.add_constant(X_train_rfe1)\nX_train_rfe1.head()\n# -\n\n# Fitting the model with 15 variables\nlm1 = sm.OLS(y_train, X_train_rfe1).fit() \nprint(lm1.summary())\n\n# Note that the new model built on the selected features doesn't show much dip in the accuracy in comparison to the model which was built on all the features. It has gone from **84.5%** to **84.4%**. This is indeed a good indication to proceed with these selected features.\n#\n# But let's check for the multicollinearity among these variables.\n\n# Check for the VIF values of the feature variables. \nfrom statsmodels.stats.outliers_influence import variance_inflation_factor\n\na=X_train_rfe1.drop('const',axis=1)\n\n# Create a dataframe that will contain the names of all the feature variables and their respective VIFs except for the constant\nvif = pd.DataFrame()\nvif['Features'] = a.columns\nvif['VIF'] = [variance_inflation_factor(a.values, i) for i in range(a.shape[1])]\nvif['VIF'] = round(vif['VIF'], 2)\nvif = vif.sort_values(by = \"VIF\", ascending = False)\nvif\n\n# +\n# RFE with 8 features\nlm = LinearRegression()\nrfe2 = RFE(lm, 8)\n\n# Fit with 7 features\nrfe2.fit(X_train, y_train)\n\n# Print the boolean results\nprint(rfe2.support_) \nprint(rfe2.ranking_) \n\n# +\n# Import statsmodels\nimport statsmodels.api as sm \n\n# Subset the features selected by rfe1\ncol1 = X_train.columns[rfe2.support_]\n\n# Subsetting training data for 7 selected columns\nX_train_rfe2 = X_train[col1]\n\n# Add a constant to the model\nX_train_rfe2 = sm.add_constant(X_train_rfe2)\nX_train_rfe2.head()\n# -\n\n# Fitting the model with 10 variables\nlm2 = sm.OLS(y_train, X_train_rfe2).fit() \nprint(lm2.summary())\n\n# Now let's check the VIF for these selected features and decide further.\n\nb=X_train_rfe2.drop('const',axis=1)\n\n# Create a dataframe that will contain the names of all the feature variables and their respective VIFs except for the constant\nvif = pd.DataFrame()\nvif['Features'] = b.columns\nvif['VIF'] = [variance_inflation_factor(b.values, i) for i in range(b.shape[1])]\nvif['VIF'] = round(vif['VIF'], 2)\nvif = vif.sort_values(by = \"VIF\", ascending = False)\nvif\n\n# From the model summary above, all the variables have p-value < 0.05 and from the p-value perspective, all variables seem significant. But notice that there are a few variables which have VIF > 5. We need to deal with these variables carefully.\n#\n# So let's try removing 'hum' first having the maximum VIF and then check for it again. Dropping this variable may result in a change in other VIFs which are high.\n\n# Let's drop the 'hum' column\nX_train_rfe2.drop(\"hum\",axis=1,inplace=True)\nX_train_rfe2\n\n# +\nX_train_rfe2 = sm.add_constant(X_train_rfe2)\n\n# Now that we have removed one variable, let's fit the model with 6 variables\nlm3 = sm.OLS(y_train, X_train_rfe2).fit() \nprint(lm3.summary())\n# -\n\n# The model seems to be doing a good job. Let's also quickly take a look at the VIF values.\n\nc=X_train_rfe2.drop('const',axis=1)\n\n# Create a dataframe that will contain the names of all the feature variables and their respective VIFs except for the constant\nvif = pd.DataFrame()\nvif['Features'] = c.columns\nvif['VIF'] = [variance_inflation_factor(c.values, i) for i in range(c.shape[1])]\nvif['VIF'] = round(vif['VIF'], 2)\nvif = vif.sort_values(by = \"VIF\", ascending = False)\nvif\n\n# All the VIF values and p-values seem to be in the permissible range now. Also the `Adjusted R-squared` value has dropped from `84.5%` with **28 variables** to just `79.3%` using **7 variables**. This model is explaining most of the variance without being too complex. So let's proceed with this model: **`lm3`**\n#\n# **Note**: Adjusted R-squared value with *9 variables* was `80.1%`, but I opted for a slightly less complex model with relatively similar accuracy but fewer features.\n\n# **So the final equation for best fitted line is:**\n#\n# `cnt` = 0.2627 + 0.2357 x yr - 0.0748 x holiday + 0.4258 x temp - 0.1519 x windspeed - 0.1407 x season_spring - 0.0726 x mnth_Jul - 0.2434 x waethersit_C\n\n# ### Residual Analysis\n#\n# Before we make predictions on the test set, let's first analyse the residuals. This is to ensure that we satisfy the underlying assumption of liner regression model.\n# - Linear relationship between X and Y - Already plotted above\n# - Error terms are normally distributed (not X, Y)\n# - Error terms have constant variance (homoscedasticity)\n# - Error terms are independent of each other\n\n# Making Prediction on train data set\ny_train_cnt = lm3.predict(X_train_rfe2)\n\ny_train_cnt.describe()\n\n# Check the Mean of Error terms\nresiduals = y_train - y_train_cnt\nmean_residuals = np.mean(residuals)\nprint(\"Mean of Residuals {}\".format(mean_residuals))\nresiduals.describe()\n\n# So, the mean of error terms is almost 0\n\n# **Check 2: Error terms are normally distributed**\n#\n# We will check if the residuals or error terms are normally distributed or not.\n\n# Plot the histogram of the error or residuals terms terms\nfig = plt.figure()\nsns.distplot((y_train - y_train_cnt), bins = 20, color = 'r')\n# Plot heading\nfig.suptitle('Error Terms', fontsize = 20) \n# Give the X-label\nplt.xlabel('Errors', fontsize = 18) \n\n# The error terms are fairly normally distributed and we can surely live with this. Let's now make predictions on the test-set.\n\n# **Check 3: Homoscedasticity**\n#\n# Homoscedasticity means that the residuals have equal or almost equal variance across the regression line. By plotting the error terms with predicted terms we should confirm that there is no pattern in the error terms.\n\nplt.figure(figsize=(15,5))\np = sns.scatterplot(y_train_cnt,residuals, color = \"r\")\nplt.xlabel('y_pred', color = 'r')\nplt.ylabel('Residuals', color = 'b')\nplt.ylim(-1,1)\nplt.xlim(0,1)\np = sns.lineplot([0,1],[0,0],color='g')\np = plt.title('Homoscedasticity Check')\nplt.show()\n\n# **Check 4. Error terms are independent of each other**\n#\n# That means there should not be any auto-correlation between error terms.\n\nplt.figure(figsize=(15,5))\np = sns.lineplot(y_train_cnt,residuals,marker='o',color='purple')\nplt.xlabel('y_pred')\nplt.ylabel('Residuals')\nplt.ylim(-1,1)\nplt.xlim(0,1)\np = sns.lineplot([0,1],[0,0],color='green')\np = plt.title('Residuals vs fitted values plot for autocorrelation check')\n\n# ### Making Predictions\n#\n# We would first need to scale the test set as well. So let's start with that.\n\nX_train_rfe2\n\n# let's recall the set of variables which are to be scaled\nvar\n\n# df_test holds the test dataset for us\ndf_test[var] = scaler.transform(df_test[var])\n\n# Split the 'df_test' set into X and y after scaling\ny_test = df_test.pop('cnt')\nX_test = df_test.drop([\"casual\",\"registered\"],axis=1)\n\nX_test.head()\n\n# Let's check the list 'col2' which had the 7 variables RFE had selected\ncol2=c.columns\nprint(len(col2))\ncol2\n\n# Let's subset these columns and create a new dataframe 'X_test_rfe1'\nX_test_rfe2 = X_test[col2]\n\n# Add a constant to the test set created\nX_test_rfe2 = sm.add_constant(X_test_rfe2)\nX_test_rfe2.info()\n\nX_test_rfe2.head()\n\n# Making predictions using our final model: lm3\ny_pred = lm3.predict(X_test_rfe2)\n\n# +\n# Plotting y_test and y_pred to understand the spread\n\nfig = plt.figure()\n#plt.scatter(y_test, y_pred)\nsns.regplot(x=y_test,y=y_pred, scatter_kws={\"color\": \"green\"}, line_kws={\"color\": \"red\"});\nfig.suptitle('y_test vs y_pred', fontsize = 16, color = 'purple') # Plot heading \nplt.xlabel('y_test', fontsize = 16) # X-label\nplt.ylabel('y_pred', fontsize = 16) # Y-label\n# -\n\n# From the above plot, it's evident that the model is doing well on the test set as well. Let's also check the R-squared and more importantly, the adjusted R-squared value for the test set.\n\n# r2_score for 8 variables on test dataset and it's predcition\nfrom sklearn.metrics import r2_score\nr2_score = r2_score(y_test, y_pred)\nprint(\"r2 score is: \",r2_score)\nn = len(X_test)\np = 7\nprint(\"and Adjusted r2 score is: \",1-(1-r2_score)*(n-1)/(n-p-1))\n\n\n# Thus, for the model with 7 variables, the r-squared on training and test data is about 79.6% and 78.35% respectively. The **Adjusted r-squared** on the train and test set is 79.3% and 77.63% respectively.\n\n# #### Checking the correlations between the final predictor variables\n\n# +\n# Figure size\nplt.figure(figsize=(8,6))\n\n# Heatmap\nsns.heatmap(bike[col2].corr(), cmap=\"YlGnBu\", annot=True)\nplt.show()\n# -\n\n# This is the simplest model that we could build. The final predictors seem to have fairly low correlations. \n#\n# Thus, the final model consists of the **7 variables** mentioned above.One can go ahead with this model and use it for predicting count of daily bike rentals.\n#\n#\n\nprint(type(lm3.params))\nprint(abs(lm3.params).sort_values(ascending = False)) # It's magnitude of the coefficinets that matters\n\n# #### Suggestion to BoomBikes Management\n#\n# We can conclude from our model that below three features are the most influential features for BIke Rentals:\n# - Temperature : with coefficient `0.42`\n# - Weather C [3-Light Snow, Light Rain + Thunderstorm + Scattered clouds, Light Rain + Scattered clouds ]: with coefficient `0.24`\n# - Year: with coefficient `0.23`\n\n# ### Conclusion: 4. Model Building and Evaluation\n#\n# In this section, we build our first model considering all the variables and then we made an informed choice based on VIF and p-values and finally made our baseline model with 7 variables. We also plotted the error term to check if they are normally distributed. Our regression line also looks pretty clustered around the central line.\n\n# ### End of assignment \n\n# **Check what's happen when we get unexpected values in any column**\n#\n# We have train our model on 2018 and 2019 data. Lets change the values of year to 2020 ( encoded as 3) and see the difference in predcited values:\n\nX_test_rfe2_O = X_test_rfe2.head(1)\nX_test_rfe2_O\n\nX_test_rfe2_C = X_test_rfe2_O.copy()\nX_test_rfe2_C\n\nX_test_rfe2_C.at[184,'yr']= 2\nX_test_rfe2_C.at[184,'weathersit_C']= 2\nX_test_rfe2_C\n\n# So, basically we changed the Year value to 2022 (coded as 2) and weathersit_C = 2 (again a wrong value because dummy variables only allow 0/1.\n#\n# Let's predict the `cnt` now for these two records seperately.\n\n# Making predictions using our final model: lm3\nprint(\"cnt for original record is:\", lm3.predict(X_test_rfe2_O))\nprint(\"cnt for changed record is:\", lm3.predict(X_test_rfe2_C))\n\n# **So Model didn't throw any error**\n\n\n","repo_name":"AmritaMLProjects/ML","sub_path":"10_Linear_Regression_Assignment/PI Session/BoomBikes - Linear Regression.ipynb","file_name":"BoomBikes - Linear Regression.ipynb","file_ext":"py","file_size_in_byte":30644,"program_lang":"python","lang":"en","doc_type":"code","stars":0,"dataset":"github-jupyter-script","pt":"40"}
+{"seq_id":"2517504603","text":"# +\nimport pandas as pd\nimport matplotlib.pyplot as plt\nimport numpy.random as np\nimport sys\nimport matplotlib\n\n# %matplotlib inline\n\n# +\n#from New York City to Key West, Florida.\ntotal = 1541\n#avg first two entries (55+65)/2\navg = 60\ndta=3\n#total date of ridden a bicycle\ntotd = 26\n\nnow=120\nkk=[55,120]\n# -\n\nlast=0\nwhile dta <= totd:\n if dta >= 26:\n last = total-now\n lastmi=now+last\n kk.append(lastmi)\n dta+=1\n else:\n now=now+60\n kk.append(now)\n dta+=1\n\nind=1\ninlis=[]\nfor i in kk:\n inlis.append(ind)\n ind+=1\n\nodometer = pd.Series(kk,index=inlis)\nodometer\n\n# +\nodometer2=[]\nlastMile=0\nday=1\nfor i in odometer:\n milePerDay=i-lastMile\n print(str(day)+'day :'+str(milePerDay)+' Mile')\n odometer2.append(milePerDay)\n day+=1\n lastMile=i\n \n\n# -\n\ndf = pd.DataFrame(data = odometer2, columns=['Mile'])\ndf.index=inlis\ndf.plot()\n","repo_name":"junphp/is362_assigment3","sub_path":"assigment3.ipynb","file_name":"assigment3.ipynb","file_ext":"py","file_size_in_byte":900,"program_lang":"python","lang":"en","doc_type":"code","stars":0,"dataset":"github-jupyter-script","pt":"40"}
+{"seq_id":"32770821099","text":"from secml.adv.attacks.poisoning import CAttackPoisoningLogisticRegression\nfrom secml.ml.classifiers import CClassifierLogistic\nfrom secml.ml.features.normalization import CNormalizerMinMax\nfrom secml.data import CDataset\nfrom secml.ml.peval.metrics import CMetricAccuracy\nimport numpy as np\nimport pandas as pd\nfrom sklearn.model_selection import train_test_split\nfrom secml.ml.classifiers import CClassifierRidge\nfrom secml.adv.attacks.poisoning import CAttackPoisoningRidge\n\n\ndata = pd.read_csv(\"sampled_darknet.csv\").values\ndatax = data[:, 1:]\ndatay = data[:, 0]\nrandom_state = 2\nx_train, x_test, y_train, y_test = train_test_split(\n datax, datay, test_size=250, train_size=1000, random_state=random_state)\nx_val, x_test, y_val, y_test = train_test_split(\n x_test, y_test, test_size=125, train_size=125, random_state=random_state)\n\n\ntraining_data = CDataset(x_train, y_train)\nvalidation_data = CDataset(x_val, y_val)\ntest_data = CDataset(x_test,y_test)\nmetric = CMetricAccuracy()\nclf=CClassifierRidge(max_iter=100)\n\n# +\nclf.fit(training_data.X, training_data.Y)\nprint(\"Training of classifier complete!\")\n# Compute predictions on a test set\ny_pred = clf.predict(test_data.X)\n\n# Bounds of the attack space. Can be set to `None` for unbounded\nlb, ub = validation_data.X.min(), validation_data.X.max()\n# Number of poisoning points to generate\nn_poisoning_points = 500\n\nsolver_params = {\n 'eta': 0.05,\n 'eta_min': 0.05,\n 'eta_max': None,\n 'max_iter': 100,\n 'eps': 1e-6\n}\n\n\n# +\npois_attack = CAttackPoisoningRidge(classifier=clf,\n training_data=training_data,\n val=validation_data,\n lb=0, ub=1,\n solver_params=solver_params,\n random_seed=random_state,\n init_type='random')\n\npois_attack.n_points = n_poisoning_points\n\n# -\n\nprint(\"Attack started...\")\npois_y_pred, pois_scores, pois_ds, f_opt = pois_attack.run(\n test_data.X, test_data.Y)\nprint(\"Attack complete!\")\n\n# +\nacc = metric.performance_score(y_true=test_data.Y, y_pred=y_pred)\n# Evaluate the accuracy after the poisoning attack\npois_acc = metric.performance_score(y_true=test_data.Y, y_pred=pois_y_pred)\n\nprint(\"Original accuracy on test set: {:.2%}\".format(acc))\nprint(\n \"Accuracy after non adaptive attack on test set: {:.2%}\".format(pois_acc))\n\n# -\n\npd.DataFrame(pois_ds.X).to_csv(\"rx_new.csv\",index=False)\ntemp=pd.read_csv(\"rx_new.csv\").values\ntemp=temp.reshape(-1,76)\npd.DataFrame(temp).to_csv(\"rx_new.csv\",index=False)\npd.DataFrame(pois_ds.Y).to_csv(\"ry_new.csv\",index=False)\n","repo_name":"dbsxfz/FBMAD","sub_path":"finaldataset/ridge.ipynb","file_name":"ridge.ipynb","file_ext":"py","file_size_in_byte":2673,"program_lang":"python","lang":"en","doc_type":"code","stars":0,"dataset":"github-jupyter-script","pt":"40"}
+{"seq_id":"20635701082","text":"# # Lesson 11\n\n# ## 00:00:07 - Blog recap\n#\n# * Blog on super convergence: [The 1cycle policy](https://sgugger.github.io/the-1cycle-policy.html)\n# * 5x faster than stepwise approaches.\n# * Let's you have massively high learning rates (somewhere between 1 and 3).\n# * Trains at high learning rates for lots of the epochs: loss doesn't improve much but it's doing a lot of searching to find generalisable areas.\n# * When learning rate is high, momentum is lower.\n# * Hamel Husain's blog on [sequence-to-sequence data products](https://towardsdatascience.com/how-to-create-data-products-that-are-magical-using-sequence-to-sequence-models-703f86a231f8)\n\n# ## 00:05:42 - Building a sequence-to-sequence model using machine translation\n#\n# * Neural translation has surpased typical translations techniques as of 2016.\n# * Path of neural translation similar to image classification in 2012: just surpased state of the art and now moving past it rapidly.\n# * Four big wins of Neural MT:\n# 1. End-to-end training: all params are optimised to minimise loss function (less hyperparams)\n# 2. Distributed rep share strength: better exploit word and phrase similarities.\n# 3. Better exploitation of context: NMT can use a much bigger context to translate more accurately.\n# 4. More fluent text generation.\n# \n# * Models use Bidirectional LSTM with Attention (which is obviously not just useful for machine translation).\n\n# ## 00:10:05 - Translate French into English\n#\n# * Basic idea: make it look like a standard NN problem, need 3 things:\n# 1. Data (x, y pairs)\n# 2. Architecture\n# 3. Loss function\n# \n# * Lots of parallel corpuses (some language -> some other language), especially for European documents.\n# * For bounding boxes, all interesting stuff is in loss function, for translation, it's all in the arch.\n#\n# ## 00:13:16 - Neural translation walkthrough\n#\n# * Take a sentence in English and put it through an RNN/Encoder.\n# * Encoder: piece of NN architecture that takes input and turns into some representation.\n# * Decoder: take the encoder / RNN output and convert into a sequence of French tokens.\n#\n# * For translating language, you don't know how many words should be outputted from a sentence.\n# * Key issue: arbitrary length output which don't correspond to the input length.\n# \n# ## 00:18:19 - RNN revision\n#\n# * Need to understand Lesson 6 if the lesson doesn't make sense.\n# * RNN is a standard fully connected network, which takes an input to a linear layer which is fed into another layer and so on. However, it has one key difference: the second layer can also accept and concat another input.\n#\n# ![](\"https://i.gyazo.com/900233717de09d0ac63b4330a2c6b877.gif\")
\n#\n# * Use the same weight matrix for each of the layer outputs and the same weight matrix for each input.\n#\n# * The diagram can be refactored to be a for loop:\n#\n# ![](\"https://i.gyazo.com/a53c737b2b3c325112430d9d3ad4b6a5.gif\")
\n#\n# ![](\"https://i.gyazo.com/82ecea54084aab349d420720a0caa647.gif\")
\n#\n# * The refactoring is basically what makes it an RNN.\n# \n# * RNNs can be stacked: output of one RNN can be fed into another:\n#\n# ![](\"https://i.gyazo.com/0383008c47d943200ea38423ffcb3071.gif\")
\n#\n# * Need to be able to write it from scratch to really understand it.\n\n# %matplotlib inline\n# %reload_ext autoreload\n# %autoreload 2\n\n# +\nimport re\nfrom pathlib import Path\nimport pickle\nfrom collections import Counter, defaultdict\n\nimport numpy as np\n\nfrom fastai.text import Tokenizer, partition_by_cores\n# -\n\nPATH = Path('data/translate')\nTMP_PATH = PATH / 'tmp'\nTMP_PATH.mkdir(exist_ok=True)\nfname = 'giga-fren.release2.fixed'\nen_fname = PATH / f'{fname}.en'\nfr_fname = PATH / f'{fname}.fr'\n\n# ### **Start do not rerun**\n\n# !wget http://www.statmt.org/wmt10/training-giga-fren.tar --directory-prefix={PATH}\n\n# !cd {PATH} && tar -xvf training-giga-fren.tar\n\n# !cd {PATH} && gunzip giga-fren.release2.fixed.en.gz && gunzip giga-fren.release2.fixed.fr.gz\n\n# * Training translation models takes a long time.\n# * No conceptual difference between 2 and 8 layers: use 2 layers because we think it should be enough.\n# * Find questions that start with Wh (what, where, when etc) and match with French questions:\n\n# +\nre_eng_questions = re.compile('^(Wh[^?.!]+\\?)')\nre_french_questions = re.compile('^([^?.!]+\\?)')\n\nen_fh = open(en_fname, encoding='utf-8')\nfr_fh = open(fr_fname, encoding='utf-8')\n\nlines = []\n\nfor eq, fq in zip(en_fh, fr_fh):\n lines.append((\n re_eng_questions.search(eq),\n re_french_questions.search(fq)\n ))\n\nquestions = [(e.group(), f.group()) for e, f in lines if e and f]\n# -\n\npickle.dump(questions, (PATH / 'fr-en-qs.pkl').open('wb'))\n\n# ### **End do not rerun**\n\nquestions = pickle.load((PATH / 'fr-en-qs.pkl').open('rb'))\n\n# * We now have 52k sentence pairs:\n\nquestions[:5], len(questions)\n\n# * Separate questions into each language:\n\nen_qs, fr_qs = zip(*questions)\n\n# ### **Start do no rerun**\n\n# * Tokenise using English and French tokenizer.\n\n# !python -m spacy download fr\n\nen_tok = Tokenizer.proc_all_mp(partition_by_cores(en_qs))\nfr_tok = Tokenizer.proc_all_mp(partition_by_cores(fr_qs), 'fr')\n\nen_tok[0], fr_tok[0]\n\n# * Want to find the largest 90th sequence sentences, and make that the max sequence length.\n\nnp.percentile([len(o) for o in en_tok], 90), np.percentile([len(o) for o in fr_tok], 90)\n\nkeep = np.array([len(o) < 30 for o in en_tok])\n\nen_tok = np.array(en_tok)[keep]\nfr_tok = np.array(fr_tok)[keep]\n\npickle.dump(en_tok, (PATH / 'en_tok.pkl').open('wb'))\npickle.dump(fr_tok, (PATH / 'fr_tok.pkl').open('wb'))\n\n# ### **End do no rerun**\n\nen_tok = pickle.load((PATH / 'en_tok.pkl').open('rb'))\nfr_tok = pickle.load((PATH / 'fr_tok.pkl').open('rb'))\n\n\n# * Don't need to know a lot of NLP stuff for deep learning on text, but the basics are useful: particurally tokenising.\n# * 00:28:37 - some students in the study group are trying to build language models for Chinese, need a tokeniser like [sentence piece](https://github.com/google/sentencepiece), since it doesn't have individual words.\n\n# * Next, turn tokens into numbers:\n\ndef toks2ids(tok, pre):\n freq = Counter(p for o in tok for p in o)\n itos = [o for o, c in freq.most_common(40000)]\n itos.insert(0, '_bos_')\n itos.insert(1, '_pad_')\n itos.insert(2, '_eos_')\n itos.insert(3, '_unk')\n stoi = defaultdict(lambda: 3, {v: k for k, v in enumerate(itos)})\n ids = np.array([([stoi[o] for o in p] + [2]) for p in tok])\n np.save(TMP_PATH / f'{pre}_ids.npy', ids)\n pickle.dump(itos, open(TMP_PATH / f'{pre}_itos.pkl', 'wb'))\n return ids, itos, stoi\n\n\nen_ids, en_itos, en_stoi = toks2ids(en_tok, 'en')\nfr_ids, fr_itos, fr_stoi = toks2ids(fr_tok, 'fr')\n\n\ndef load_ids(pre):\n ids = np.load(TMP_PATH / f'{pre}_ids.npy')\n itos = pickle.load(open(TMP_PATH / f'{pre}_itos.pkl', 'rb'))\n stoi = defaultdict(lambda: 3, {v:k for k, v in enumerate(itos)})\n return ids, itos, stoi\n\n\nen_ids, en_itos, en_stoi = load_ids('en')\nfr_ids, fr_itos, fr_stoi = load_ids('fr')\n\n[fr_itos[o] for o in fr_ids[0]], len(en_itos), len(fr_itos)\n\n# ## 00:33:01 - Word vectors\n\n# * Seq-to-seq with language models hasn't been explored yet in academia: lots of potential papers to be written.\n# * Word2Vec has been surpased by a number of word vectors: FastText is a good choice.\n\n# #### **Start do not rerun**\n\n# !pip install git+https://github.com/facebookresearch/fastText.git\n\n# * Need to also download the fasttext word vectors:\n\n# +\n# #!wget https://s3-us-west-1.amazonaws.com/fasttext-vectors/wiki.en.zip --directory-prefix={PATH}\n\n# +\n# #!wget https://s3-us-west-1.amazonaws.com/fasttext-vectors/wiki.fr.zip --directory-prefix={PATH}\n# -\n\n# !cd {PATH} && unzip wiki.en.zip && unzip wiki.fr.zip\n\n# #### **End do not rerun**\n\nimport fastText as ft\n\nenglish_vecs = ft.load_model(str((PATH / 'wiki.en.bin')))\n\nfrench_vecs = ft.load_model(str((PATH / 'wiki.fr.bin')))\n\n\n# * Turn it into a dictionary:\n\ndef get_vecs(lang, ft_vecs):\n vecd = {w: ft_vecs.get_word_vector(w) for w in ft_vecs.get_words()}\n pickle.dump(vecd, open(PATH / 'wiki.{lang}.pkl'))\n\n\n# * Preparing data for PyTorch\n\nen_ids_tr = np.array([o[:enlen_90] for o in en_ids])\nfr_ids_tr = np.array([o[:rnlen_90] for o in fr_ids])\nenlen_90, frlen_90\n\n\ndef iter_to_numpy(*a):\n \"\"\"convert iterable object into numpy array\"\"\"\n return np.array(a[0]) if len(a)==1 else [np.array(o) for o in a]\n\n\nclass Seq2SeqDataset(Dataset):\n def __init__(self, x, y):\n self.x = x\n self.y = y\n \n def __getitem__(self, idx):\n return iter_to_numpy(self.x[idx], self.y[idx])\n \n def __len__(self):\n return len(self.x)\n\n\n\nnp.random.seed(42)\ntrn_keep = np.random.rand(len(en_ids_tr)) > 0.1\nen_trn, fr_trn = es_ids_tr[trn_keep], fr_ids_tr[trn_keep]\nen_val, fr_val = en_ids_tr[~trn_keep], fr_ids_tr[~trn_keep]\nlen(en_trn), len(en_val)\n\ntrn_ds = Seq2SeqDataset(fr_trn, en_trn)\nval_ds = Seq2SeqDataset(fr_val, en_val)\n\nbs = 125\n\n\n","repo_name":"lextoumbourou/notes","sub_path":"notes/reference/moocs/fast.ai/dl-2018/lesson11.ipynb","file_name":"lesson11.ipynb","file_ext":"py","file_size_in_byte":9020,"program_lang":"python","lang":"en","doc_type":"code","stars":52,"dataset":"github-jupyter-script","pt":"40"}
+{"seq_id":"18358287905","text":"# +\nimport numpy as np\nimport pandas as pd\n\ntemp = pd.read_csv(\"C://Users/ARPIT/Desktop/titanic/titanic_train.csv\")\ntemp\n# -\n\ntemp.head()\n\ntemp.describe()\n\ntemp2 = temp.copy()\ntemp2.head()\n\ndel temp2[\"Name\"]\n\n\ntemp2.head()\n\n# +\ndel temp2[\"Ticket\"]\n\ntemp2.head()\n\n\n# +\ndef a(n):\n if n == \"male\":\n return 1\n else:\n return 0\n\n\ntemp2[\"gender\"] = temp2.Sex.apply(a)\n\ntemp2.head(9)\n\n# -\n\ndel temp2[\"Sex\"]\ntemp2.head()\n\ntemp2.isnull().sum()\n\ntemp2.Cabin.fillna(0,inplace = True)\ntemp2.head()\n\na = temp2[temp2.Survived==1]\na\n\ntemp2.Age.fillna(a.Age.mean(),inplace = True)\ntemp2\n\n# +\nb = temp2[temp2.Survived==0]\n\ntemp2.Age.fillna(b.Age.mean(),inplace = True)\n\ntemp2\n# -\n\ntemp2.isnull().sum()\n\n# temp2.Embarkment\n\ntemp2.Embarked\n\n\n# +\ndef e(n):\n if n == 'S':\n return 1\n else:\n return 0\n\n\ntemp2[\"southhampton\"] = temp2.Embarked.apply(e)\n\ntemp2.head()\n\n\n# +\ndef c(n):\n if n == 'C':\n return 1\n else:\n return 0\n\n\ntemp2[\"cherbourgh\"] = temp2.Embarked.apply(c)\n\ntemp2.head()\n\n\n# +\ndef q(n):\n if n == 'Q':\n return 1\n else:\n return 0\n\n\ntemp2[\"queenstown\"] = temp2.Embarked.apply(q)\n\ntemp2.head()\n# -\n\ndel temp2[\"Embarked\"]\ntemp2.head()\n\ntemp2.isnull().sum()\n\ntemp2.describe()\n\ntemp2.to_csv(\"C://Users/ARPIT/Desktop/titanic/titanic_clean.csv\")\n\n\n","repo_name":"danzra/titanic-dataset-analysis","sub_path":"titanic_cleaned.ipynb","file_name":"titanic_cleaned.ipynb","file_ext":"py","file_size_in_byte":1308,"program_lang":"python","lang":"en","doc_type":"code","stars":0,"dataset":"github-jupyter-script","pt":"40"}
+{"seq_id":"17296095670","text":"# + [markdown] id=\"view-in-github\" colab_type=\"text\"\n# \n\n# + id=\"SPxxjzOWj43w\"\nfrom keras.datasets import fashion_mnist\nimport matplotlib.pyplot as plt\n# %matplotlib inline\nfrom sklearn.model_selection import train_test_split\nfrom keras.utils import np_utils\n\n# + id=\"YAjFHa5nj5vg\"\n(X_train, y_train), (X_test, y_test) = fashion_mnist.load_data() # loading the fashion mnist dataset\n\n# + colab={\"base_uri\": \"https://localhost:8080/\"} id=\"JDlguyEwj7xe\" outputId=\"7fff81ac-e3d3-45b7-a210-b1b9ec1f8edb\"\nprint(X_train.shape, y_train.shape)\nprint(X_test.shape, y_test.shape)\n\n# + colab={\"base_uri\": \"https://localhost:8080/\"} id=\"N_QRSk5akGOd\" outputId=\"f998a3ef-e950-4014-9814-3595c1229164\"\nX_train[0]\n\n# + colab={\"base_uri\": \"https://localhost:8080/\"} id=\"KWqJUt_lkPHA\" outputId=\"4749e944-6f90-4cf3-b12c-2b0d18009dee\"\ny_train[0]\n\n# + colab={\"base_uri\": \"https://localhost:8080/\", \"height\": 265} id=\"Xteg3HjAkQhQ\" outputId=\"2774647b-b7b0-42a6-fb74-baf80efe5a6b\"\nplt.imshow(X_train[0], cmap='gray')\nplt.show()\n\n# + id=\"9lip_v2lkqQt\"\nX_train, X_valid, y_train, y_valid = train_test_split(X_train, y_train, test_size=0.2, random_state=42) \n\n# + id=\"MWZ83h3_lOfb\"\nX_train = X_train.astype('float32') / 255 # normalization so that the values lie between 0 and 1\nX_valid = X_valid.astype('float32') / 255\n\n# + id=\"i1k27RbzlfgZ\"\nX_test = X_test.astype('float32') / 255\n\n# + colab={\"base_uri\": \"https://localhost:8080/\"} id=\"SiBQP7T4lqj9\" outputId=\"3f1d611b-f87a-430e-9796-6edd14be7490\"\nX_train.shape, X_valid.shape, X_test.shape\n\n# + id=\"ESeRskf2lutS\"\nX_train = X_train.reshape(-1, 28, 28, 1) # reshaping it ti (-1,28,28,1) because it's a gray scale image and it has only one channel\nX_valid = X_valid.reshape(-1, 28, 28, 1)\nX_test = X_test.reshape(-1, 28, 28, 1)\n\n# + colab={\"base_uri\": \"https://localhost:8080/\"} id=\"wQk2RGxhl_ED\" outputId=\"134fc253-9236-403f-b194-a31aa5180f10\"\nX_train.shape, X_valid.shape, X_test.shape\n\n# + id=\"BVxAq1nUmCMX\"\ny_train = np_utils.to_categorical(y_train, num_classes=10) # converting the outpt to categorical form. ie: 5 = [0,0,0,0,0,1,0,0,0,0], 3 = [0,0,0,1,0,0,0,0,0,0]\ny_valid = np_utils.to_categorical(y_valid, num_classes=10)\ny_test = np_utils.to_categorical(y_test, num_classes=10)\n\n# + colab={\"base_uri\": \"https://localhost:8080/\"} id=\"QaP05umuntML\" outputId=\"07d1fee0-19ca-4a8d-94f3-5ce0ddd03c7a\"\ny_train.shape, y_valid.shape, y_test.shape\n\n# + id=\"1GdG2f3locNf\"\nfrom keras.models import Sequential, load_model\nfrom keras.layers import Conv2D, MaxPool2D, Flatten, Dense, Dropout\n\n# + id=\"QfyzigKKpaiL\"\nmodel = Sequential([ # initializing model as a sequential model\n Conv2D(filters=64, kernel_size=7, input_shape=[28, 28, 1], activation='relu', padding='same'), # conv2D layer to pick the necessary features\n MaxPool2D(pool_size=2), # MaxPool2D to down sample so that it prevents over fitting\n Conv2D(filters = 128, kernel_size=3, activation='relu', padding='same'), # padding to make sure the dimensionality remains the same\n Conv2D(filters = 128, kernel_size=3, activation='relu', padding='same'), # relu for activation\n MaxPool2D(pool_size=2),\n Conv2D(filters = 256, kernel_size=3, activation='relu', padding='same'),\n Conv2D(filters = 256, kernel_size=3, activation='relu', padding='same'),\n MaxPool2D(pool_size=2),\n Flatten(), # flatten to make it a 1D array\n Dense(128, activation='relu'), # dense layer for classification \n Dropout(0.5), # dropout layer to prevent over-fitting\n Dense(64, activation='relu'),\n Dropout(0.5),\n Dense(10, activation='softmax') # softmax because we have a categorical output\n])\n\n# + colab={\"base_uri\": \"https://localhost:8080/\"} id=\"HNbBnR-hpdR2\" outputId=\"fb2f21d3-58cb-487d-80a9-4c17f6cecc97\"\nmodel.summary()\n\n# + id=\"gsmuyTckraeu\"\nmodel.compile(loss='categorical_crossentropy', optimizer='nadam', metrics=['accuracy'])\n\n# + id=\"hSetsyLlrhSe\"\nfrom keras.callbacks import ModelCheckpoint\n\n# + id=\"wQKBYoPLsJ2u\"\ncb = ModelCheckpoint('try_1', monitor='val_accuracy', save_best_only=True)\n\n# + colab={\"base_uri\": \"https://localhost:8080/\"} id=\"C6VhRPANtU12\" outputId=\"4eb3f129-8476-455d-eac1-3f50630993c4\"\ntrain = history = model.fit(X_train, y_train, epochs=30, validation_data=(X_valid, y_valid), callbacks=[cb])\n\n# + colab={\"base_uri\": \"https://localhost:8080/\"} id=\"kzkuq9YZuCHO\" outputId=\"bdb5d0be-9dbd-41b6-985c-3891fe11d7f0\"\nmodel = load_model('try_1')\nmodel.evaluate(X_test, y_test)\n\n# + id=\"zs5MtwjjzeBG\"\n\n","repo_name":"nikhil2312/Fashion-mnist","sub_path":"fashion_mnist.ipynb","file_name":"fashion_mnist.ipynb","file_ext":"py","file_size_in_byte":4973,"program_lang":"python","lang":"en","doc_type":"code","stars":0,"dataset":"github-jupyter-script","pt":"40"}
+{"seq_id":"14328339765","text":"import generax as gx\nimport jax\nimport jax.numpy as jnp\nimport jax.random as random\nimport matplotlib.pyplot as plt\nfrom generax.training.trainer import Trainer\nimport generax.util as util\nfrom functools import partial\n\n# +\n# Get the dataset\nfrom sklearn.datasets import make_moons, make_swiss_roll\ndata, y = make_swiss_roll(n_samples=100000, noise=0.5)\ndata = data[:, [0, 2]]\ndata = data - data.mean(axis=0)\ndata = data/data.std(axis=0)\nkey = random.PRNGKey(0)\n\ndef get_train_ds(key, batch_size: int = 128):\n total_choices = jnp.arange(data.shape[0])\n closed_over_data = data # In case we change the variable \"data\"\n while True:\n key, _ = random.split(key, 2)\n idx = random.choice(key,\n total_choices,\n shape=(batch_size,),\n replace=True)\n yield dict(x=closed_over_data[idx])\n\ntrain_ds = get_train_ds(key)\n\ntrue_samples = util.extract_multiple_batches_from_iterator(train_ds,\n n_batches=10,\n single_batch=True)\nplt.scatter(*true_samples['x'].T)\n\n# +\nx = data[:10]\n\nfrom generax.distributions.flow_models import RealNVP, NeuralSpline, ContinuousNormalizingFlow\n\nmodel = ContinuousNormalizingFlow(input_shape=x.shape[1:],\n key=key,\n n_blocks=5,\n hidden_size=32,\n working_size=16,\n time_embedding_size=32,\n n_time_features=4*32,\n controller_atol=1e-7,\n controller_rtol=1e-7)\n\n# +\nfrom generax.training import FlowMatching\nimport optax\n\n# Create the optimizer\nschedule = optax.warmup_cosine_decay_schedule(init_value=0.0,\n peak_value=1.0,\n warmup_steps=1000,\n decay_steps=3e5,\n end_value=0.1,\n exponent=1.0)\nchain = []\nchain.append(optax.clip_by_global_norm(15.0))\nchain.append(optax.adamw(1e-3))\nchain.append(optax.scale_by_schedule(schedule))\noptimizer = optax.chain(*chain)\n\n# Create the loss function\nfm = FlowMatching(path_type='straight',\n coupling_type='ot')\n\n# Data dependent initialization for the vector field\ndata = next(train_ds)\nmodel = fm.initialize_vector_field(model, data, key)\n\n# Create the trainer\ntrainer = Trainer(checkpoint_path='tmp/flow_matching')\nmodel = trainer.train(model=model,\n objective=fm.loss_function,\n evaluate_model=lambda x: x,\n optimizer=optimizer,\n num_steps=20000,\n double_batch=1000,\n data_iterator=train_ds,\n checkpoint_every=1000,\n test_every=-1,\n retrain=False)\n# -\n\nsamples = model.sample(key, n_samples=1000)\n\n# +\nT = 5\nts = jnp.linspace(0.0, 1.0, T)\n\ndef get_time_samples(z):\n log_pz = model.prior.log_prob(z)\n x, log_det = model.transform.neural_ode(z,\n inverse=True,\n log_likelihood=True,\n save_at=ts)\n log_px = log_pz - log_det\n return x, log_px\n\nz = random.normal(key, (10000, 2))\nsamples, log_px = jax.vmap(get_time_samples)(z)\nsamples.shape\n# -\n\nlog_px.shape\n\n# +\nn_rows, n_cols = 1, T\nsize = 4\nfig, axes = plt.subplots(1, T, figsize=(n_cols*size, n_rows*size))\nax_iter = iter(axes.ravel())\nfor i, t in enumerate(ts):\n ax = next(ax_iter)\n ax.scatter(*samples[:, i].T, alpha=0.7, s=3, c=jnp.exp(log_px)[:, i])\n ax.set_title(f't={t:3.2f}')\n ax.set_axis_off()\n\nimport os\n# Get the path for ~\nhome = os.path.expanduser(\"~\")\nsave_dir = os.path.join(home, 'eddiecunningham.github.io', 'content', 'images')\nsave_path = os.path.join(save_dir, 'cnf_evolution.png')\nplt.savefig(save_path, bbox_inches='tight')\n# -\n\n\n\n\n","repo_name":"EddieCunningham/EddieCunningham.github.io","sub_path":"notebooks/cnf.ipynb","file_name":"cnf.ipynb","file_ext":"py","file_size_in_byte":4092,"program_lang":"python","lang":"en","doc_type":"code","stars":0,"dataset":"github-jupyter-script","pt":"41"}
+{"seq_id":"4483292237","text":"import os\nimport numpy as np\nimport pickle\nimport matplotlib.pyplot as plt\nfrom preprocessing_helper import *\n\n# ## 1. Load in preprocessed data\n\ncwd = os.getcwd()\nparent_wd = cwd.replace('/preprocessing', '')\nurl_data_path = parent_wd + '/raw_data/moonGen_scrape_2016_cp'\n\nwith open(url_data_path, 'rb') as f:\n MoonBoard_2016_withurl = pickle.load(f)\n\nX_path_all = [cwd + '/benchmark_withgrade_move_seq_X', \n cwd + '/benchmark_nograde_move_seq_X', \n cwd + '/nonbenchmark_withgrade_move_seq_X', \n cwd + '/nonbenchmark_nograde_move_seq_X']\nY_path_all = [cwd + '/benchmark_withgrade_move_seq_Y', \n cwd + '/benchmark_nograde_move_seq_Y', \n cwd + '/nonbenchmark_withgrade_move_seq_Y', \n cwd + '/nonbenchmark_nograde_move_seq_Y']\n\nX_seq_dict_merge = {}\nY_seq_dict_merge = {}\nfor path in X_path_all:\n with open(path, 'rb') as f:\n read_dict = pickle.load(f)\n X_seq_dict_merge = {**X_seq_dict_merge, **read_dict}\nfor path in Y_path_all:\n with open(path, 'rb') as f:\n read_dict = pickle.load(f)\n Y_seq_dict_merge = {**Y_seq_dict_merge, **read_dict}\n\n# ### 1-1. remove data from fail list and save\n\nfail_list = ['363336', \n '363335', \n '350368', \n '349610', \n '349049', \n '348915', \n '348858', \n '348670', \n '348669', \n '348432', \n '346738', \n '344743', \n '339916', \n '339325', \n '337916', \n '335566', \n '312004', \n '310949', \n '309657', \n '309311', \n '248997', \n '246092', \n '231401', \n '231392', \n '19362', \n '360322', \n '356219', \n '322560', \n '311585', \n '309230', \n '308089']\n\nfor key in fail_list:\n try:\n del X_seq_dict_merge[key]\n except:\n pass\n try:\n del Y_seq_dict_merge[key]\n except:\n pass\nassert len(X_seq_dict_merge) == len(Y_seq_dict_merge)\n\nsave_pickle(X_seq_dict_merge, cwd + '/X_seq_dict_merge')\nsave_pickle(Y_seq_dict_merge, cwd + '/Y_seq_dict_merge')\n\n# ## 2. Partition to train/dev/test set\n\n# ### 2-1. convert to matrix form\n\n# +\nn_sample = len(Y_seq_dict_merge)\nX_seq_data_merge = np.zeros((n_sample, 12, 22))\nY_seq_grade_merge = np.zeros(n_sample)\nkeys_seq_merge = []\ntmax_seq_merge = np.zeros(n_sample)\n\ni = 0\nfor key, value in X_seq_dict_merge.items():\n X_data = value.T\n X_seq_data_merge[i, 0:X_data.shape[0], :] = X_data\n Y_seq_grade_merge[i] = Y_seq_dict_merge[key]\n keys_seq_merge.append(key)\n tmax_seq_merge[i] = X_data.shape[0]\n i = i + 1\n# -\n\nplt.figure(dpi = 150)\nresult = plt.hist(Y_seq_grade_merge, bins = np.arange(11)-0.5)\nfor x,y in zip(np.arange(10),result[0]):\n label = str(int(y))\n plt.annotate(label,\n (x,y), \n textcoords=\"offset points\",\n xytext=(0,3),\n ha='center')\nplt.xticks(np.arange(10), ['V4','V5','V6','V7','V8','V9','V10','V11','V12','V13'])\nplt.xlabel('Grade')\nplt.ylabel('number count')\nplt.show()\n\n# ### 2-2. Partition\n\nn_dev = 3000\nn_test = 3000\nn_train = n_sample - n_dev - n_test\nprint(n_train)\n\nnp.random.seed(0)\nshuffle = np.random.choice(np.arange(n_sample), n_sample, replace = False)\n\nX_seq_shuffle = X_seq_data_merge[shuffle, :, :]\nY_seq_shuffle = Y_seq_grade_merge[shuffle]\nkeys_seq_shuffle = np.array(keys_seq_merge)[shuffle]\ntmax_seq_shuffle = tmax_seq_merge[shuffle]\n\ntraining_set_seq = {'X': X_seq_shuffle[0:n_train], \n 'Y': Y_seq_shuffle[0:n_train], \n 'keys': keys_seq_shuffle[0:n_train], \n 'tmax': tmax_seq_shuffle[0:n_train]}\ndev_set_seq = {'X': X_seq_shuffle[n_train:n_train+n_dev], \n 'Y': Y_seq_shuffle[n_train:n_train+n_dev], \n 'keys': keys_seq_shuffle[n_train:n_train+n_dev], \n 'tmax': tmax_seq_shuffle[n_train:n_train+n_dev]}\ntest_set_seq = {'X': X_seq_shuffle[n_train+n_dev:], \n 'Y': Y_seq_shuffle[n_train+n_dev:], \n 'keys': keys_seq_shuffle[n_train+n_dev:], \n 'tmax': tmax_seq_shuffle[n_train+n_dev:]}\n\n# +\n# training_set_path = parent_wd + '/preprocessing/training_set_seq_12'\n# dev_set_path = parent_wd + '/preprocessing/dev_set_seq_12'\n# test_set_path = parent_wd + '/preprocessing/test_set_seq_12'\n# save_pickle(training_set_seq, training_set_path)\n# save_pickle(dev_set_seq, dev_set_path)\n# save_pickle(test_set_seq, test_set_path)\n# -\n\n# ### 2-3. remove problem in dev & test sets that has 0 repeat\n\n# +\ndel_list = []\nfor key in training_set_seq['keys']:\n if MoonBoard_2016_withurl[key]['repeats'] == 0:\n del_list.append(key)\n\ndel_list_devtest = []\nfor key in dev_set_seq['keys']:\n if MoonBoard_2016_withurl[key]['repeats'] == 0:\n del_list_devtest.append(key)\nfor key in test_set_seq['keys']:\n if MoonBoard_2016_withurl[key]['repeats'] == 0:\n del_list_devtest.append(key)\n\n# +\nrow_rm = []\nfor i, key in enumerate(training_set_seq['keys']):\n if key in del_list:\n row_rm.append(i)\n\ntraining_set_seq['X'] = np.delete(training_set_seq['X'], row_rm, 0)\ntraining_set_seq['Y'] = np.delete(training_set_seq['Y'], row_rm)\ntraining_set_seq['keys'] = np.delete(training_set_seq['keys'], row_rm)\ntraining_set_seq['tmax'] = np.delete(training_set_seq['tmax'], row_rm)\n\n# +\nrow_rm_dev = []\nfor i, key in enumerate(dev_set_seq['keys']):\n if key in del_list_devtest:\n row_rm_dev.append(i)\n\ndev_set_seq['X'] = np.delete(dev_set_seq['X'], row_rm_dev, 0)\ndev_set_seq['Y'] = np.delete(dev_set_seq['Y'], row_rm_dev)\ndev_set_seq['keys'] = np.delete(dev_set_seq['keys'], row_rm_dev)\ndev_set_seq['tmax'] = np.delete(dev_set_seq['tmax'], row_rm_dev)\n\nrow_rm_test = []\nfor i, key in enumerate(test_set_seq['keys']):\n if key in del_list_devtest:\n row_rm_test.append(i)\n\ntest_set_seq['X'] = np.delete(test_set_seq['X'], row_rm_test, 0)\ntest_set_seq['Y'] = np.delete(test_set_seq['Y'], row_rm_test)\ntest_set_seq['keys'] = np.delete(test_set_seq['keys'], row_rm_test)\ntest_set_seq['tmax'] = np.delete(test_set_seq['tmax'], row_rm_test)\n# -\n\n# #### Visualize the new distribution of data after removing 0 repeat problems\n\nplt.figure(dpi = 150)\nresult = plt.hist(np.hstack([training_set_seq['Y'], dev_set_seq['Y'], test_set_seq['Y']]), bins = np.arange(11)-0.5)\nfor x,y in zip(np.arange(10),result[0]):\n label = str(int(y))\n plt.annotate(label,\n (x,y),\n textcoords=\"offset points\",\n xytext=(0,3),\n ha='center')\nplt.xticks(np.arange(10), ['V4','V5','V6','V7','V8','V9','V10','V11','V12','V13'])\nplt.xlabel('Grade')\nplt.ylabel('number count')\nplt.show()\n\n# +\n# training_set_path = parent_wd + '/preprocessing/training_set_seq_12_rmrp0'\n# dev_set_path = parent_wd + '/preprocessing/dev_set_seq_12_rmrp0'\n# test_set_path = parent_wd + '/preprocessing/test_set_seq_12_rmrp0'\n# save_pickle(training_set_seq, training_set_path)\n# save_pickle(dev_set_seq, dev_set_path)\n# save_pickle(test_set_seq, test_set_path)\n# -\n\n# ### 2-4. normalization\n\ntraining_seq_normalized = normalization(training_set_seq)\ndev_seq_normalized = normalization(dev_set_seq)\ntest_seq_normalized = normalization(test_set_seq)\n\n# ### 2-5. save\n\ntraining_normalized_set_path = parent_wd + '/preprocessing/training_seq_n_12_rmrp0'\ndev_normalized_set_path = parent_wd + '/preprocessing/dev_seq_n_12_rmrp0'\ntest_normalized_set_path = parent_wd + '/preprocessing/test_seq_n_12_rmrp0'\nsave_pickle(training_seq_normalized, training_normalized_set_path)\nsave_pickle(dev_seq_normalized, dev_normalized_set_path)\nsave_pickle(test_seq_normalized, test_normalized_set_path)\n","repo_name":"jrchang612/MoonBoardRNN","sub_path":"preprocessing/Step3_partition_train_test_set_v2.ipynb","file_name":"Step3_partition_train_test_set_v2.ipynb","file_ext":"py","file_size_in_byte":7829,"program_lang":"python","lang":"en","doc_type":"code","stars":23,"dataset":"github-jupyter-script","pt":"41"}
+{"seq_id":"8196038309","text":"# # JOIN\n#\n# This script joins the following datasets, all using pandas:\n# - Schools\n# - Wind \n# - PM2.5 readings\n# - Population\n# - Closest pollution sources for each school\n#\n# \n\n# + colab={\"base_uri\": \"https://localhost:8080/\", \"height\": 542} id=\"3kDBwAcJRscD\" outputId=\"ae3461b7-9c6b-41e0-cb10-43e627d14e04\"\n# Q1. Load the \"titanic\" dataset using the load_dataset function of seaborn. Use Plotly express to plot a scatter plot for age and fare columns in the titanic dataset.\n\nimport seaborn as sns\nimport plotly.express as px\n\n# Load the \"titanic\" dataset\ntitanic_data = sns.load_dataset(\"titanic\")\n\n# Plot a scatter plot for age and fare columns using Plotly express\nscatter_fig = px.scatter(titanic_data, x=\"age\", y=\"fare\", title=\"Scatter Plot of Age and Fare in Titanic Dataset\")\nscatter_fig.show()\n\n# + colab={\"base_uri\": \"https://localhost:8080/\", \"height\": 542} id=\"x99iGqOBSFKi\" outputId=\"355b711c-6cd0-4c4e-922f-b3f1a92bf7b1\"\n# Q2. Using the tips dataset in the Plotly library, plot a box plot using Plotly express.\n\n\nimport plotly.express as px\n\n# Load the \"tips\" dataset from Plotly library\ntips_data = px.data.tips()\n\n# Plot a box plot using Plotly express\nbox_fig = px.box(tips_data, x=\"day\", y=\"total_bill\", title=\"Box Plot of Total Bill for Different Days\")\nbox_fig.show()\n\n\n# + colab={\"base_uri\": \"https://localhost:8080/\", \"height\": 542} id=\"un473iW-SJqI\" outputId=\"1acb9586-152b-47a3-fad1-a256b5d20c02\"\n# Q3. Using the tips dataset in the Plotly library, Plot a histogram for x=\"sex\" and y=\"total_bill\" column in the tips dataset. Also, use the \"smoker\" column with the pattern_shape parameter and the \"day\" column with the color parameter.\n\nimport plotly.express as px\n\n# Load the \"tips\" dataset from Plotly library\ntips_data = px.data.tips()\n\n# Plot a histogram using Plotly express\nhistogram_fig = px.histogram(tips_data, x=\"sex\", y=\"total_bill\", color=\"day\", barmode=\"group\",\n pattern_shape=\"smoker\", title=\"Histogram of Total Bill by Sex and Day\")\nhistogram_fig.show()\n\n\n\n# + colab={\"base_uri\": \"https://localhost:8080/\", \"height\": 542} id=\"eqgsZXbMSOw4\" outputId=\"7facce31-1b4a-4fc3-f553-b06b82e47e60\"\n# Q4. Using the iris dataset in the Plotly library, Plot a scatter matrix plot, using the \"species\" column for the color parameter. Note: Use \"sepal_length\", \"sepal_width\", \"petal_length\", \"petal_width\" columns only with the dimensions parameter.\n\nimport plotly.express as px\n\n# Load the \"iris\" dataset from Plotly library\niris_data = px.data.iris()\n\n# Plot a scatter matrix plot using Plotly express\nscatter_matrix_fig = px.scatter_matrix(iris_data, dimensions=[\"sepal_length\", \"sepal_width\", \"petal_length\", \"petal_width\"],\n color=\"species\", title=\"Scatter Matrix Plot of Iris Dataset\")\nscatter_matrix_fig.show()\n\n# + colab={\"base_uri\": \"https://localhost:8080/\", \"height\": 542} id=\"WUIbfS7KSahN\" outputId=\"7490c776-5a30-411d-92cc-78c03916ed0f\"\n# Q5. What is Distplot? Using Plotly express, plot a distplot.\n\n\"\"\"Ans 5: `Distplot` is a function from the seaborn library that combines a histogram and a kernel density plot to visualize the distribution of a single numerical variable.\nThe histogram represents the frequency distribution of the data, while the kernel density plot provides a smooth estimate of the data's probability density.\n\"\"\"\nimport seaborn as sns\nimport plotly.express as px\n\n# Load sample data\ntips_data = sns.load_dataset(\"tips\")\n\n# Plot a distplot using Plotly express\ndistplot_fig = px.histogram(tips_data, x=\"total_bill\", marginal=\"rug\", title=\"Distplot of Total Bill\")\ndistplot_fig.show()\n","repo_name":"afzalasar7/Data-Science","sub_path":"Week 9 Visualization Tools/Plotly.ipynb","file_name":"Plotly.ipynb","file_ext":"py","file_size_in_byte":3728,"program_lang":"python","lang":"en","doc_type":"code","stars":1,"dataset":"github-jupyter-script","pt":"41"}
+{"seq_id":"10119234979","text":"# # Overview\n#\n# This code demonstrates the NeuralNetwork class by training a neural network to recognize handwritten digits (the MNIST dataset)\n\n# +\n# Plot entropy loss\nplt.plot(1-np.asarray(entropyNN.training_accuracy), label=\"Training Error\")\nplt.plot(np.asarray(entropyNN.validation_accuracy), label=\"Validation Accuracy\")\nplt.ylabel(\"Percent\")\nplt.xlabel(\"Iteration Number(Thousands)\")\nplt.title(\"Training Error and Validation Accuracy for Entropy Loss\")\nlegend = plt.legend(loc=\"best\")\nplt.show()\n\n\n# -\n\n# Plot entropy loss\nplt.plot(1-np.asarray(entropyNN.training_accuracy[0:9]), label=\"Training Error\")\nplt.plot(np.asarray(entropyNN.validation_accuracy[0:9]), label=\"Validation Accuracy\")\nplt.ylabel(\"Percent\")\nplt.xlabel(\"Iteration Number(Ten Thousands)\")\nplt.title(\"Training Error and Validation Accuracy for Entropy Loss (First 10,000 iterations in Batches of 100)\")\nlegend = plt.legend(loc=\"best\")\nplt.show()\n\n# Plot squared loss\nplt.plot(1-np.asarray(squareNN.training_accuracy), label=\"Training Error\")\nplt.plot(np.asarray(squareNN.training_accuracy), label=\"Validation Accuracy\")\nlegend = plt.legend(loc=\"best\")\nplt.ylabel(\"Percent\")\nplt.xlabel(\"Iteration Number(Thousands)\")\nplt.title(\"Training Error and Validation Accuracy for Squared Loss\")\nplt.show\n\n# Plot for batch size of 100, ran from 10,000x iterations\n# Plot squared loss\nplt.plot(1-np.asarray(squareNN.training_accuracy[0:9]), label=\"Training Error\")\nplt.plot(np.asarray(squareNN.validation_accuracy[0:9]), label=\"Validation Accuracy\")\nlegend = plt.legend(loc=\"best\")\nplt.ylabel(\"Percent\")\nplt.xlabel(\"Iteration Number(Ten Thousands)\")\nplt.title(\"Training Error and Validation Accuracy for Squared Loss (First 10,000 iterations in Batches of 100)\")\nplt.show\n\n\nentropyNN = NeuralNetwork(784, 200, 10, \"entropy\")\nentropyNN.train(Xtrain, Ytrain, Xvalid, Yvalid, 10000, 10)\n\nsquareNN = NeuralNetwork(\"square\")\nsquareNN.train(Xtrain, Ytrain, Xvalid, Yvalid, 10000, 10)\n\nentropyNN.train(Xtrain, Ytrain, Xvalid, Yvalid, 250000, 100)\n\nsquareNN.train(Xtrain, Ytrain, Xvalid, Yvalid, 250000, 100)\n\n# Implementation notes are as follows:\n#\n# Learning rate: The learning rate is initialized to be 0.01 and then decreased by 10 percent every 1000 iterations. When training is subsequently called again, the training rate is reinitialized to 0.01\n#\n# Stopping condition: The change in the training accuracy rate is calculated per iteration, and if the absolute change in the accuracy is less than 1E-5, then the training stops. For Kaggle, I kept on calling the training function for as much time as I had before the submission deadline. \n#\n# Weights: The weights were initialized to be drawn from a uniform random distribution between 0 and 0.01. \n#\n# Training set and Validation set: The data was randomly shuffled in place to save memory and then 59000 samples were separated into the training set and 1000 samples were separated into the validation set. \n#\n# Training error and validation accuracy: Graphs of training error and validation accuracy are shown below. The neural network trained on entropy loss converged faster than the neural network trained on squared loss, with an asymptotic curve versus the squared loss' more linear improvement. After 10000 iterations, the entropy loss resulted in lower training error of 0.127 vs the squared loss' training error of 0.462. Likewise, the entropy loss' validation accuracy of 0.881 is higher than the squared loss' validation accuracy of 0.438. After 10000 iterations of 100 points, entropy has a better training accuracy. To summarize, the neural network trained on entropy is better, because it's training error decreases faster and the validation accuracy increases faster than the neural network trained on squared loss\n#\n# Running time: See python output. Entropy is slightly slower. I probably trained it for several hours in total. Around the 0.9 percent accuracy rate, the score increassed really slowly. Prediction score: \\textbf{0.9668}.\n#\n# Extra features: I implemented batch gradient descent with a user specified number of desired samples per batch. I used batch sizes of 10 due to concern about running out of memory or needing to stop training. \n#\n","repo_name":"LilyHu/neuralnetwork","sub_path":"NeuralNetwork-RecognizingHandwrittenDigitsDemo.ipynb","file_name":"NeuralNetwork-RecognizingHandwrittenDigitsDemo.ipynb","file_ext":"py","file_size_in_byte":4185,"program_lang":"python","lang":"en","doc_type":"code","stars":0,"dataset":"github-jupyter-script","pt":"41"}
+{"seq_id":"15443966153","text":"# + [markdown] id=\"view-in-github\" colab_type=\"text\"\n# \n\n# + [markdown] id=\"DE4HRAFRGH0g\"\n# # **Next Word Suggestion**\n#\n# ### **Overview**\n# The task of Next Word Suggestion comes under the NLP Task: Masked Language Modelling. The model gets an input with a mask:\n# **Sample sentence: **\n#\n# we are [MASK].\n#\n# we are eating.\n#\n# we are dancing.\n#\n# we are playing.\n#\n#\n# ### **Methodology**\n# In this project, we will use pre-trained models from hugging face and fine tune it on different dataset. We will use the model \"BERT-UNCASED\".\n#\n# ### **AI Application**\n#\n#\n# * NLP\n#\n# ### **Business Segments**\n#\n#\n# * Lifestyle & Social Media, Media & Publishing\n# * Business & Private Sector\n#\n# ### **Data**\n# [Dataset link](https://huggingface.co/datasets/empathetic_dialogues)\n#\n#\n#\n#\n#\n#\n#\n#\n#\n#\n#\n\n# + colab={\"base_uri\": \"https://localhost:8080/\"} id=\"Jmxrc3d8GbcQ\" outputId=\"4639a3d2-2267-4843-ffc5-c7c34adbf35a\"\n# !pip install transformers\n\n# + colab={\"base_uri\": \"https://localhost:8080/\"} id=\"gZWF31e1GxoC\" outputId=\"1a189cbb-f318-4b0d-b5bc-4a01961d0944\"\n# !pip install datasets transformers accelerate\n\n\n# + [markdown] id=\"XBRd92G8HWd3\"\n# ## **IMPORT LIBRARIES**\n\n# + id=\"U7IU8CdcGH_y\"\nfrom transformers import AutoModelForMaskedLM\nfrom transformers import AutoTokenizer\nfrom datasets import load_dataset\nfrom transformers import DataCollatorForLanguageModeling\nfrom transformers import TrainingArguments\nfrom torch.utils.data import DataLoader\nfrom transformers import default_data_collator\nfrom torch.optim import AdamW\nfrom accelerate import Accelerator\nfrom transformers import get_scheduler\nfrom tqdm.auto import tqdm\nimport torch\nimport math\nfrom transformers import pipeline\n\n\n# + [markdown] id=\"5YjTlWd7Hbqd\"\n# ## **LOAD THE DATASET**\n\n# + colab={\"base_uri\": \"https://localhost:8080/\"} id=\"nFidJG1UGTol\" outputId=\"aad476f3-cf35-4e4d-9743-979215df6a7a\"\ndataset = load_dataset(\"empathetic_dialogues\", split='train')\n\n# + [markdown] id=\"CrrB-5T7Hi2C\"\n# ## **PRE-PROCESSING THE DATA**\n\n# + id=\"shSSpfqEHI1g\"\n# Renaming and dropping unrequired columns\ndataset = dataset.remove_columns([col for col in dataset.column_names if col != 'utterance'])\ndataset = dataset.rename_column('utterance', 'text')\n\n# + [markdown] id=\"8pzCDa3nIEGv\"\n# ## **MODEL**\n#\n#\n\n# + [markdown] id=\"ExYtqpI_ITTR\"\n# # **BERT-BASE-UNCASED** (A pre-trained model from hugging face)\n\n# + colab={\"base_uri\": \"https://localhost:8080/\"} id=\"FpN5nVwzIBG5\" outputId=\"1b70eafb-561f-4950-fbdd-8ec89f161b05\"\nmodel_checkpoint = \"bert-base-uncased\"\nmodel = AutoModelForMaskedLM.from_pretrained(model_checkpoint)\n\n# + colab={\"base_uri\": \"https://localhost:8080/\"} id=\"FSQ6ggkTIcDJ\" outputId=\"1e16f588-2018-43ce-f4e0-9a70b79b0a32\"\nbert_num_parameters = model.num_parameters() / 1_000_000\nprint(f\"'>>> BERT number of parameters: {round(bert_num_parameters)}M'\")\n\n# + id=\"SXlldzM2Ityf\"\ntext = \"I love [MASK].\"\n\n# + [markdown] id=\"k9HyJO9UJFy1\"\n# ## **Importing Auto-tokenizer**\n\n# + id=\"DXK7pirsJInn\"\ntokenizer = AutoTokenizer.from_pretrained(model_checkpoint)\n\n\n# + id=\"gZVwryrNJLeX\"\ndef tokenize_function(examples):\n result = tokenizer(examples[\"text\"], max_length=512,\n padding='max_length', truncation=True, return_tensors=\"pt\")\n if tokenizer.is_fast:\n result[\"word_ids\"] = [result.word_ids(i) for i in range(len(result[\"input_ids\"]))]\n return result\n\n\n# + colab={\"base_uri\": \"https://localhost:8080/\"} id=\"Eu2rgDMgJW8i\" outputId=\"6658ad43-f0c2-46f4-a0e2-ef5acc1b8df8\"\ntokenized_datasets = dataset.map(\n tokenize_function, batched=True, remove_columns=[\"text\"]\n)\ntokenized_datasets\n\n# + colab={\"base_uri\": \"https://localhost:8080/\"} id=\"OuROzNENJwNQ\" outputId=\"84d566f5-56de-4ed7-ac9f-c0d15431b0ca\"\ntokenizer.model_max_length\n\n# + [markdown] id=\"xGl510WLKGIx\"\n# ## **Creating Chunk Of Data**\n\n# + id=\"OcJp3f_YKA4U\"\nchunk_size = 128\n\n\n# + colab={\"base_uri\": \"https://localhost:8080/\"} id=\"Tq7sXNXZKOpM\" outputId=\"e3936820-caf3-4b04-ea21-80bcf61e5210\"\ndef group_texts(examples):\n # Concatenate all texts\n concatenated_examples = {k: sum(examples[k], []) for k in examples.keys()}\n # Compute length of concatenated texts\n total_length = len(concatenated_examples[list(examples.keys())[0]])\n # We drop the last chunk if it's smaller than chunk_size\n total_length = (total_length // chunk_size) * chunk_size\n # Split by chunks of max_len\n result = {\n k: [t[i : i + chunk_size] for i in range(0, total_length, chunk_size)]\n for k, t in concatenated_examples.items()\n }\n # Create a new labels column\n result[\"labels\"] = result[\"input_ids\"].copy()\n return result\n\nlm_datasets = tokenized_datasets.map(group_texts, batched=True)\nlm_datasets\n\n# + [markdown] id=\"RtTUCofdL4rI\"\n# ## **Downsampling the dataset using train_test_split**\n\n# + colab={\"base_uri\": \"https://localhost:8080/\"} id=\"ZpdIym68KRJI\" outputId=\"f43bf53e-e544-4650-fb8b-cffcb4a977df\"\ntrain_size = 10_000\ntest_size = int(0.1 * train_size)\n\ndownsampled_dataset = lm_datasets.train_test_split(\n train_size=train_size, test_size=test_size, seed=42\n)\ndownsampled_dataset\n\n# + id=\"BPFtrSY7L8hT\"\ndata_collator = DataCollatorForLanguageModeling(tokenizer=tokenizer, mlm_probability=0.15)\n\n# + [markdown] id=\"nZc-WYZRMOj7\"\n# ## **Training the Argument from transformer**\n\n# + id=\"teFP7ivlZjdN\"\nbatch_size = 32\n# Show the training loss with every epoch\nlogging_steps = len(downsampled_dataset[\"train\"]) // batch_size\nmodel_name = model_checkpoint.split(\"/\")[-1]\n\ntraining_args = TrainingArguments(\n output_dir=f\"{model_name}-finetuned-empathetic dialogues\",\n overwrite_output_dir=True,\n evaluation_strategy=\"epoch\",\n learning_rate=2e-5,\n weight_decay=0.01,\n per_device_train_batch_size=batch_size,\n per_device_eval_batch_size=batch_size,\n push_to_hub=True,\n fp16=True,\n logging_steps=logging_steps,\n)\n\n\n# + id=\"-SVoYnYGZrOn\"\ndef insert_random_mask(batch):\n features = [dict(zip(batch, t)) for t in zip(*batch.values())]\n masked_inputs = data_collator(features)\n # Create a new \"masked\" column for each column in the dataset\n return {\"masked_\" + k: v.numpy() for k, v in masked_inputs.items()}\n\n\n# + colab={\"base_uri\": \"https://localhost:8080/\", \"height\": 34, \"referenced_widgets\": [\"d6ce0775f1b3449282d84579606abc50\", \"4bf92ce6d59d423bb6425aceafa7178f\", \"90e171e7f1a74d00b701d92946d5dcf9\", \"855f652675c84dab8094db76312d471a\", \"7d40ab67f743418b9a88ffe6d363f6b1\", \"88bb7a71ae1a42b4815e97c26b07dca4\", \"7f14a71ee573457ba855476a74e925bb\", \"530deb30f12a49d8afa4c9056678bd34\", \"3727f3799e444e0ea98520e9963486d6\", \"5087d7f192554397932e28099a42bbad\", \"921bda81bebc4cf784259884bba806d9\"]} id=\"17_sTJYwax1i\" outputId=\"b1ddcff9-6b11-4006-f665-7ffbc1d6713c\"\ndownsampled_dataset = downsampled_dataset.remove_columns([\"word_ids\"])\neval_dataset = downsampled_dataset[\"test\"].map(\n insert_random_mask,\n batched=True,\n remove_columns=downsampled_dataset[\"test\"].column_names,\n)\neval_dataset = eval_dataset.rename_columns(\n {\n \"masked_input_ids\": \"input_ids\",\n \"masked_attention_mask\": \"attention_mask\",\n \"masked_labels\": \"labels\",\n }\n)\n\n# + [markdown] id=\"Zi-O0BycbAX9\"\n# ## **Training the model**\n\n# + id=\"dmeWmBiaa10f\"\nbatch_size = 32\ntrain_dataloader = DataLoader(\n downsampled_dataset[\"train\"],\n shuffle=True,\n batch_size=batch_size,\n collate_fn=data_collator,\n)\neval_dataloader = DataLoader(\n eval_dataset, batch_size=batch_size, collate_fn=default_data_collator\n)\n\n# + id=\"mlwhw0Rra_mF\"\noptimizer = AdamW(model.parameters(), lr=5e-5)\n\n# + [markdown] id=\"bAQDQzZSbOst\"\n# ## **Training With Accelerator**\n\n# + id=\"8anq7tNdbG8N\"\naccelerator = Accelerator()\nmodel, optimizer, train_dataloader, eval_dataloader = accelerator.prepare(\n model, optimizer, train_dataloader, eval_dataloader\n)\n\n# + [markdown] id=\"JiM1fz0ubYJ2\"\n# ## **Scheduling Learning Rate**\n\n# + id=\"XgHwJ51EbL-P\"\nnum_train_epochs = 3\nnum_update_steps_per_epoch = len(train_dataloader)\nnum_training_steps = num_train_epochs * num_update_steps_per_epoch\n\nlr_scheduler = get_scheduler(\n \"linear\",\n optimizer=optimizer,\n num_warmup_steps=0,\n num_training_steps=num_training_steps,\n)\n\n# + id=\"z6uvN5xybWLN\"\noutput_dir = f\"bert-base-uncased-finetuned-empathetic dialogues\"\n\n# + colab={\"base_uri\": \"https://localhost:8080/\", \"height\": 101, \"referenced_widgets\": [\"cb0ef830b6db4af8ac1a79007ed6873c\", \"e39563b3a5bf4157abcd1865b9520dfa\", \"565ac4785dcc404e9846d1fc10730249\", \"c1784f2383db4e98989f06efaeb08ff1\", \"6902496d85564c9d98362d5fd112f309\", \"a6abcad9d5a44ee2b392c4c19438fe33\", \"45acbbcc8096467994a73e57d9530d7a\", \"761852cd186d49be80893968434e39e4\", \"a1944e29ed68425caf61dd932fe6b0ed\", \"0f902aae9bb44d02ba372ff7ee79c757\", \"7ec9bdd157114301bf437a8db5a8e15a\"]} id=\"Qeaa6JbAblYM\" outputId=\"f517ff8a-a9cb-41b6-adad-ae57e542b75c\"\nprogress_bar = tqdm(range(num_training_steps))\n\nfor epoch in range(num_train_epochs):\n # Training\n model.train()\n for batch in train_dataloader:\n # # Remove the masked_token_type_ids argument\n inputs = {k: v for k, v in batch.items() if k != \"masked_token_type_ids\"}\n outputs = model(**batch)\n loss = outputs.loss\n accelerator.backward(loss)\n\n optimizer.step()\n lr_scheduler.step()\n optimizer.zero_grad()\n progress_bar.update(1)\n\n # Evaluation\n model.eval()\n losses = []\n for step, batch in enumerate(eval_dataloader):\n # Remove the masked_token_type_ids argument\n inputs = {k: v for k, v in batch.items() if k != \"masked_token_type_ids\"}\n with torch.no_grad():\n # Use the updated inputs dictionary\n outputs = model(**inputs)\n\n loss = outputs.loss\n losses.append(accelerator.gather(loss.repeat(batch_size)))\n\n losses = torch.cat(losses)\n losses = losses[: len(eval_dataset)]\n try:\n perplexity = math.exp(torch.mean(losses))\n except OverflowError:\n perplexity = float(\"inf\")\n\n print(f\">>> Epoch {epoch}: Perplexity: {perplexity}\")\n\n # Save and upload\n accelerator.wait_for_everyone()\n unwrapped_model = accelerator.unwrap_model(model)\n unwrapped_model.save_pretrained(output_dir, save_function=accelerator.save)\n if accelerator.is_main_process:\n tokenizer.save_pretrained(output_dir)\n\n\n# + [markdown] id=\"ha9dAyaYiOW3\"\n# ## **Making Predictions on the MASK.**\n\n# + id=\"TuYzI8escBX0\"\nmask_filler = pipeline(\n \"fill-mask\", model=\"/content/bert-base-uncased-finetuned-empathetic dialogues\"\n)\n\n\n# + id=\"smiQHdjrfdSi\"\ndef predict(sample=''):\n preds = mask_filler(sample)\n df = pd.DataFrame(preds)\n df['score'] = df['score'].apply(lambda x: round(x, 2))\n print(f'Sample Text: {sample}')\n print('-'*50)\n print('Model Predictions')\n print('-'*50)\n print(df)\n print('-'*50)\n return df\n\n\n# + id=\"EVtU751ihB_U\"\nimport pandas as pd\n\n# + colab={\"base_uri\": \"https://localhost:8080/\"} id=\"nKSLjrtEgZno\" outputId=\"1ec2e9e9-eb42-460b-d04b-cdc7e5ca4c95\"\ndf = predict(sample='I am [MASK] to the shop.')\n\n# + colab={\"base_uri\": \"https://localhost:8080/\"} id=\"Oz_Cz_qUg48g\" outputId=\"baf7d4a7-232f-44fe-98d6-a947b90ddc8e\"\ndf2 = predict(sample='I am [MASK] in the park.')\n","repo_name":"priyanka011011/NextWordSuggestion","sub_path":"Bert_empathetic_dialogues.ipynb","file_name":"Bert_empathetic_dialogues.ipynb","file_ext":"py","file_size_in_byte":11423,"program_lang":"python","lang":"en","doc_type":"code","stars":0,"dataset":"github-jupyter-script","pt":"41"}
+{"seq_id":"10363551449","text":"# # E-commerce website 'Wish' Summer Clothes Sale EDA\n\nimport numpy as np\nimport pandas as pd\nimport matplotlib.pyplot as plt\n# %matplotlib inline\nimport seaborn as sns\nimport pandasql as ps\n\n# ## Reading the dataset\n\ndata = pd.read_csv(r\"C:\\Users\\Twinkle Sahni\\Downloads\\jfreex-summer-products-and-sales-performance-in-e-commerce-on-wish\\jfreex-summer-products-and-sales-performance-in-e-commerce-on-wish\\summer-products-with-rating-and-performance_2020-08.csv\")\n\ndata.head()\n\nprint(\"Data shape :\",data.shape)\n\ndata.info()\n\n# ## Data Preprocessing\n\n# ### Dropping the irrelevant features\n\ndf = data.copy()\n\ndf.drop(['currency_buyer', 'badges_count', 'badge_local_product', 'badge_product_quality', 'badge_fast_shipping',\n 'product_variation_inventory','has_urgency_banner', 'urgency_text',\n 'merchant_title','merchant_info_subtitle','merchant_has_profile_picture', \n 'merchant_profile_picture', 'product_url', 'product_picture', 'theme', 'crawl_month'], axis = 1, inplace = True)\n\n# ### Cleaning the data\n\ndf.isnull().sum()\n\n# The rating count has zero null values while individual numerical rating count has 45 nulls, implying that for 45 products we do not have segmented ratings which will not be an issue in our analysis. Similarly, 41 items has no information about the color. We can drop these values from our dataset.\n# Moreover, for 17 merchandise, the country for origin is not known and I think in this case, the null values can be filled using Central Tendency Measures, i.e., Mode. Missing merchant names can just be dopped off the dataset.\n\ndf['origin_country'].mode().values\n\ndf['origin_country'].replace('NaN', 'CN')\n\ndf = df.dropna()\n\nnull1 = df.isnull().sum().to_frame(name='nulls').T\nnull1\n\n# ## Analyzing the data\n\ndf.head(10)\n\ndf.describe().T\n\ndf.dtypes\n\ncorr = df.corr()\ncorr\n\nfig, ax = plt.subplots(figsize=(12,8))\nfig = sns.heatmap(df.corr(), annot=True, cmap=sns.cubehelix_palette(as_cmap=True), ax=ax)\n\n# As we can see from the heatmap, the correlation of our target variable, units sold, is the highest (0.9) with number of rating of a product followed by count of rating of a merchant. On the contrary, the correlation between units sold and price is found to be not so strong.\n\n# ### Units Sold\n\n# Getting the distinct values of units_sold to group them for better analysis.\n\nq1= \"\"\"\nSELECT DISTINCT units_sold\nFROM df\norder by units_sold\n\"\"\"\nquery1 = ps.sqldf(q1, locals())\nquery1\n\ndf['units_sold_group'] = df['units_sold'].apply(lambda x: 1 if x <= 100 else 2 if x == 1000 else 3 if x == 5000 else 4 \\\n if x == 10000 else 5 if x == 20000 else 6 if x == 50000 else 7)\ndf.units_sold_group\n\n# Groups of units_sold:\n# 1. Less than 100\n# 2. Between 101 and 1000\n# 3. Between 1001 and 5000\n# 4. Between 5001 and 10000\n# 5. Between 10001 and 20000\n# 6. Between 20001 and 50000\n# 7. Between 50001 and 100000\n\n# #### Discount\n\n# Due to lack of complete information, we will consider the price as discounted price while retail price is the maximum retail price. So, the difference between the two can be accounted a discount given on each product during the summer sale.\n\ndf['discount'] = df['retail_price'] - df['price']\ndf['discount_percn'] = round(df['discount']/df['retail_price'] * 100, 2)\nprint(df['discount_percn'].head(10))\n\ncorr_sale_discount = df['units_sold'].corr(df['discount'])\ncorr_sale_discount\n\n# The discount feature appears to be only weakly positively correlated with number of units sold.\n\n# +\nfig, ax = plt.subplots(figsize = (15, 10))\nsns.barplot(x = df.units_sold_group, y = df.discount_percn, ax = ax)\n\nunits_sold_order = ['0.1k-1k', '1k-5k', '5k-10k', '10k-20k', '20k-50k', '50k-100k', '100k+']\nax.set_xticklabels(units_sold_order)\nax.set_ylabel('Discount (%)', fontweight = 'bold')\nax.set_xlabel('Units Sold', fontweight = 'bold')\n# -\n\n# We can conclude from the above diagram that higher the discount, higher is the number of clothes sold during the sale as this has been also confirmed by the positive correlation between the two. 100k+ units have a discount of nearly 65% while number of products selling between 1k-50k having a discount range of 20%-30%. This notion confirms the price sensitive behaviour of buyers.\n\n# #### User Ratings\n\n# In this section, we look at 5 most rated and 5 least rated products along with graphical relationship between ratings and number of products sold.\n\n# +\nq2 = \"\"\"\nSELECT title as Product_Name, rating_count as Ratings\nfrom df\norder by Ratings desc\nlimit 5 \n\"\"\"\n\nquery2 = ps.sqldf(q2, locals())\npd.set_option('display.max_colwidth', None)\nquery2\n\n# +\nq3 = \"\"\"\nSELECT title as Product_Name, rating_count as Ratings\nfrom df\norder by Ratings\nlimit 5\n\"\"\"\n\nquery3 = ps.sqldf(q3, locals())\nquery3\n\n# +\nfig, ax = plt.subplots(figsize = (15, 10))\nsns.barplot(x = df.units_sold_group, y = df.rating_count, ax = ax)\n\nunits_sold_order = ['0.1k-1k', '1k-5k', '5k-10k', '10k-20k', '20k-50k', '50k-100k', '100k+']\nax.set_xticklabels(units_sold_order)\nax.set_ylabel('Rating Count', fontweight = 'bold')\nax.set_xlabel('Units Sold', fontweight = 'bold')\n# -\n\n# #### Ad Boosts\n\nq4 = \"\"\"\nSELECT count(units_sold) as Sale_with_adboosts\nfrom df\nwhere uses_ad_boosts = 1\nUNION\nSELECT count(units_sold) as Sale_without_adboosts\nfrom df\nwhere uses_ad_boosts = 0\n\"\"\"\nquery4 = ps.sqldf(q4, locals())\nquery4\n\n# +\nfig, ax = plt.subplots(figsize = (10, 12))\nsns.barplot(x = 'units_sold_group', y='price', data=df, hue='uses_ad_boosts')\n\nunits_sold_order = ['0.1k-1k', '1k-5k', '5k-10k', '10k-20k', '20k-50k', '50k-100k', '100k+']\nax.set_xticklabels(units_sold_order)\nax.set_ylabel('Price (Euro)', fontweight = 'bold')\nax.set_xlabel('Units Sold', fontweight = 'bold')\n# -\n\n# Insight: Ad boost has no significant impact on the sale of the products. This is a signal for the marketing team work on better and compelling campaigns. \n\n# ### Product Features\n\n# #### Size\n\nq5 = \"\"\"\nSELECT distinct(product_variation_size_id)\nfrom df\n\"\"\"\nquery5 = ps.sqldf(q5, locals())\nquery5['product_variation_size_id'] = query5['product_variation_size_id'].apply(lambda x: x.upper())\nquery5 = list(query5['product_variation_size_id'])\nquery5\n\n# +\nxxs_list = []\nxs_list = []\ns_list = []\nm_list = []\nl_list = []\nxl_list = []\nxxl_list = []\n\nfor i in query5:\n if (i == 'XXS') or (i == 'XXXS') or (i == 'SIZE XXS') or (i == 'SIZE-XXS') or (i == 'SIZE -XXS') or (i=='XS.') or (i == 'SizeXS.'):\n xxs_list.append(i)\n elif (i == 'XS') or (i == 'XS.') or (i == 'SIZE XS') or (i == 'SIZE-XS'):\n xs_list.append(i)\n elif (i == 'S') or (i == 'S.') or (i == 'SUIT-S') or (i == 'SIZE S.') or (i == 'SIZE S') or (i == 'S..') \\\n or (i == 'S(BUST 88CM)') or (i == 'S (WAIST58-62CM)') or (i == 'SIZE-S') or (i == 'S DIAMETER 30CM') or (i == '25-S') \\\n or (i == 'SIZE/S') or (i == 'PANTS-S') or (i == 'SIZE--S'):\n s_list.append(i)\n elif (i == 'M') or (i == 'M.') or (i == 'SIZE M'):\n m_list.append(i)\n elif (i == 'L') or (i == '32/L') or (i == 'L.') or (i == 'SIZEL'):\n l_list.append(i)\n elif (i == 'XL') or (i == 'X L'):\n xl_list.append(i)\n elif (i == 'XXL') or (i == 'XXXXXL') or (i == '3XL') or (i == 'SIZE-4XL') or (i == 'XXXL') or (i == '5XL') \\\n or (i == '2XL') or (i == '4XL') or (i == 'XXXXL') or (i == '6XL') or (i == 'SIZE-5XL') or (i == 'SIZE4XL') \\\n or (i == '04-3XL') or (i == '1 PC - XL'):\n xxl_list.append(i)\n \ndef size(tipe):\n if tipe in xxs_list:\n return 'XXS'\n elif tipe in xs_list:\n return 'XS'\n elif tipe in s_list:\n return 'S'\n elif tipe in m_list:\n return 'M'\n elif tipe in l_list:\n return 'L'\n elif tipe in xl_list:\n return 'XL'\n elif tipe in xxl_list:\n return 'XXL'\n else:\n return 'OTHER'\n \ndf['product_variation_size_id'] = df['product_variation_size_id'].apply(size) \n\nProduct_size = pd.DataFrame(df['product_variation_size_id'].value_counts())\nProduct_size['product_size_per'] = Product_size['product_variation_size_id']/Product_size['product_variation_size_id'].sum()\nProduct_size\n# -\n\nfig = plt.figure(figsize = (10, 8))\nplt.bar(Product_size.index, Product_size.product_variation_size_id)\nplt.xlabel(\"Sizes\")\nplt.ylabel(\"Number of products sold\")\nplt.title(\"Number of units old by sizes\")\nplt.show()\n\n# Insight: S is the most sold size followed by XS and M during the summer sale.\n\n# #### Color\n\nq6 = \"\"\"\nSELECT distinct product_color\nfrom df\n\"\"\"\nquery6 = ps.sqldf(q6, locals())\nquery6 = list(query6['product_color'])\nquery6\n\n# +\nred_list = []\ngreen_list = []\nwhite_list = []\nblack_list = []\nblue_list = []\nyellow_list = []\ngrey_list = []\npurple_list = []\n\nfor i in query6:\n if (i == 'red') or (i == 'RED') or (i == 'wine red') or (i == 'rose red') or (i == \"lightred\") or (i == 'coralred') \\\n or (i == 'watermelonred') and (i != 'white & red') and (i != 'red & blue'):\n red_list.append(i)\n elif (i == 'green') or (i == 'applegreen') or (i == 'light green') or (i == 'mintgreen') or (i == 'lightgreen') \\\n or (i == 'armygreen') and (i != 'white & green') and (i != 'black & green'):\n green_list.append(i)\n elif (i == 'white') or (i == 'White') or (i == 'whitefloral') or (i == 'offwhite') or (i == 'whitestripe'):\n white_list.append(i)\n elif (i == 'black')or (i == 'Black') or (i == 'Coolblack') or (i == 'Offblack'):\n black_list.append(i)\n elif (i == 'blue') or (i == 'Blue') or (i == ' navyblue') or (i == 'navy blue') or (i == 'lightblue') or (i == 'lakeblue') and \\\n (i != 'navyblue & white') and (i != 'black & blue') and (i != 'blue & pink') and (i != 'pink & blue'):\n blue_list.append(i)\n elif (i == 'Yellow') or (i == 'Lightyellow'):\n yellow_list.append(i)\n elif (i == 'Grey') or (i == 'Greysnakeskinprint') or (i == 'Lightgrey') or (i == 'Gray') or (i == 'Lightgray'):\n grey_list.append(i)\n elif 'Purple' in i:\n purple_list.append(i)\n \ndef color(tipe):\n if tipe in red_list:\n return 'Red'\n elif tipe in green_list:\n return 'Green'\n elif tipe in white_list:\n return 'White'\n elif tipe in black_list:\n return 'Black'\n elif tipe in blue_list:\n return 'Blue'\n elif tipe in yellow_list:\n return 'Yellow'\n elif tipe in grey_list:\n return 'Gray'\n elif tipe in purple_list:\n return 'Purple'\n else:\n return 'Other'\n \ndf['product_color'] = df['product_color'].apply(color)\nProduct_color = pd.DataFrame(df['product_color'].value_counts())\nProduct_color['product_color_per'] = Product_color['product_color']/Product_color['product_color'].sum()\nProduct_color\n# -\n\nfig = plt.figure(figsize = (10, 8))\nplt.bar(Product_color.index, Product_color.product_color)\nplt.xlabel(\"Colors\")\nplt.ylabel(\"Number of products sold\")\nplt.title(\"Number of units old by colors\")\nplt.show()\n\n# Insight: From the sorted data, Black is the most demanded color (19.7%) followed by White (16.4%). The category of 'Other' colors include colors like Pink, Maroon, Gold, Brown etc.\n\n# #### Gender\n\n# +\ndf['title_orig'] = df['title_orig'].apply(lambda x: x.title())\n\nman_list = []\n\ndef prod_gender(tipe):\n for i in df['title_orig']:\n if ('Man' in i) or ('Men' in i) and (i != 'Woman') :\n man_list.append(i)\n if tipe in man_list:\n return 'Man'\n else:\n return 'Woman'\n\n\ndf['pro_gender'] = df['title_orig'].apply(prod_gender)\ndf['pro_gender'].value_counts()\n# -\n\n# Insight: Roughly 94% of all the products constitute women clothes.\n\n# ### Merchant Rating\n\n# +\nq7 = \"\"\"\nSELECT merchant_name, merchant_rating_count, merchant_rating, merchant_id \nfrom df\norder by merchant_rating_count desc \nlimit 5\n\"\"\"\n\nquery7 = ps.sqldf(q7, locals())\nquery7\n\n# +\nq8 = \"\"\"\nSELECT title, units_sold\nfrom df\nwhere merchant_name = \"simplevalueltd\"\norder by units_sold\n\"\"\"\n\nquery8 = ps.sqldf(q8, locals())\nquery8\n# -\n\n# Insight: Merchant 'simplevalueltd' is the highest rated merchant with a rating of 4.35 and more than 2 million counts. Their top selling product is a Cat imprint shirt for women.\n\n# +\nfig, ax = plt.subplots(figsize = (15, 10))\nsns.barplot(x = df.units_sold_group, y = df.merchant_rating_count, ax = ax)\n\nunits_sold_order = ['0.1k-1k', '1k-5k', '5k-10k', '10k-20k', '20k-50k', '50k-100k', '100k+']\nax.set_xticklabels(units_sold_order)\nax.set_ylabel('Number of Merchant Ratings', fontweight = 'bold')\nax.set_xlabel('Units Sold', fontweight = 'bold')\n# -\n\n# Higher the number of Ratings of a Merchant, higher is the units sold for that merchant.\n\n# +\nfig, ax = plt.subplots(figsize = (15, 10))\nsns.barplot(x = df.units_sold_group, y = df.merchant_rating, ax = ax)\n\nunits_sold_order = ['0.1k-1k', '1k-5k', '5k-10k', '10k-20k', '20k-50k', '50k-100k', '100k+']\nax.set_xticklabels(units_sold_order)\nax.set_ylabel('Number of Merchant Ratings', fontweight = 'bold')\nax.set_xlabel('Units Sold', fontweight = 'bold')\n# -\n\n# Products sold are the ones with merchant_rating of 4 or more. This is a indication of buyer awareness implying that customers tend to buy products from higher rated merchants only.\n\n# +\nq9 = \"\"\"\nSELECT title as Product_Name, units_sold as Units_Sold\nfrom df\norder by units_sold desc\nlimit 5 \n\"\"\"\n\nquery9 = ps.sqldf(q9, locals())\npd.set_option('display.max_colwidth', None)\nquery9\n# -\n\n\n","repo_name":"Twinklesahni23/EDA-using-python-and-SQL","sub_path":"Wish_Ecommerce_EDA.ipynb","file_name":"Wish_Ecommerce_EDA.ipynb","file_ext":"py","file_size_in_byte":13301,"program_lang":"python","lang":"en","doc_type":"code","stars":0,"dataset":"github-jupyter-script","pt":"41"}
+{"seq_id":"32411071117","text":"word=[[['紅色','橘色'],'黃色','綠色'],'藍色','紫色']\nword[2]\n\nword[0][2]\n\nword[0][0][1]\n\nword2=['a','b','c','d','e','f']\n\nword2[0]='A'\nword2\n\nword2[0:2]\n\nword2[::2]\n\nword2[::-1]\n\nword2.append('g')\nword2\n\nword3=['1','2','3','4','5','6']\nword2.extend(word3)\nword2\n\nword2.insert(0,'123')\nword2\n\ndel word2[-1]\nword2\n\nword2.remove('123')\nword2\n\nword2.index('e')\n\n'H' in word2\n\n'a' in word2\n\n'A' in word2\n\nword2.count('1')\n\nword2.count('0')\n\n','.join(word2)\n\nseparator='*'\njoined=separator.join(word2)\njoined\n\nseparated=joined.split(separator)\nseparated\n\nseparated==word2\n\nword4=['3','5','1','2','4','9']\nword4.sort()\nword4\n\nword4.sort(reverse=True)\nword4\n\nlen(word4)\n\na=['1','2','3']\na\n\nb=a\nb\n\na[0]='ww'\na\n\nb\n\nb[0]='aa'\nb\n\na\n\na=['1','2','3']\nb=a.copy()\nc=list(a)\nd=a[:]\na[0]='123'\na\n\nb\n\nc\n\nd\n\nword5=(1,2,3,4,5)\na,b,c,d,e=word5\nc\n\nd\n\ne\n\nword6=('+','-','*','/')\nword6\n\nword5,word6=word6,word5\nword5\n\nword6\n\n\n","repo_name":"HueiYou/B10409037","sub_path":"Chapter3_2.ipynb","file_name":"Chapter3_2.ipynb","file_ext":"py","file_size_in_byte":913,"program_lang":"python","lang":"en","doc_type":"code","stars":0,"dataset":"github-jupyter-script","pt":"41"}
+{"seq_id":"14411016993","text":"class Human:\n def __init__(self, name, age, sex):\n self.name = name\n self.age = age\n self.sex = sex\n print(\"응애응애\")\n def who(self):\n print(self.name, self.age, self.sex)\n def setinfo(self, name, age, sex):\n self.name = name\n self.age = age\n self.sex = sex\n def __del__(self):\n print(\"죽음\")\n\n\n\ntom = Human()\n\nareum = Human(\"아름\", 25, \"여자\")\n\nareum.age\n\nareum.who()\n\ntom = Human(\"모름\",0,\"모름\")\n\ntom.setinfo(\"탐\", 26, \"남\")\n\ntom.age\n\ndel tom\n\n\n# https://wikidocs.net/7036\n","repo_name":"yungsuKo/pythonStudy","sub_path":"300example_for_beginner/11. 파이썬 클래스.ipynb","file_name":"11. 파이썬 클래스.ipynb","file_ext":"py","file_size_in_byte":568,"program_lang":"python","lang":"en","doc_type":"code","stars":0,"dataset":"github-jupyter-script","pt":"41"}
+{"seq_id":"21942653848","text":"# # Statistiques\n\n# La classe statistiques permet d'abord de collecter la population par un grand nombre de variable pour chaque année. Avec ce cube de données, la classe permet ensuite de sortir les statistiques de nombres et proportions. On en donne un exemple ici. \n\n# ## Simulation scénario de référence\n\nimport sys\nsys.path.append('/Users/ydecarie/Dropbox (CEDIA)/OLG_CAN/simgen/')\nfrom matplotlib import pyplot as plt\nimport warnings\nwarnings.filterwarnings(\"ignore\")\n\nfrom simgen import model\n\nyr_debut=2017\nyr_fin=2060\n\nbase = model(start_yr=yr_debut,stop_yr=yr_fin)\nbase.startpop('startpopchsld')\nbase.immig_assumptions(init='newimmpopchsld')\nbase.birth_assumptions(scenario='reference')\nbase.dead_assumptions(scenario='low')\n\nbase.simulate(rep=1)\n\nbase.stats.save('simpop.pkl')\n\n# ## Le cube de donnée\n#\n# À l'intérieur d'un modèle, une instance de la classe stats a été lancée. Elle définie des stratas pour cumuler les fréquences d'individus. Par défaut, elle fait (age,male,educ,married,nkids). Après avoir crée l'instance, le modèle va l'initatiser en utilisant la population de départ. Il va ensuite la populer avec les années subséquentes de la simulation. On peut voir ce cube à l'aide de stats.counts. \n\nbase.stats.counts\n\n# ## Population par âge et totale\n\n# On peut obtenir la population totale en utilisant la fonction\n\nbase.stats.freq()\n\n# Et un beau graphique en rajoutant plot()\n\nbase.stats.freq().plot()\n\n# On voit qu'on a un problème au départ. Pour diagnostiquer, allons voir la population par âge: \n\npopage = base.stats.freq('age')\npopage.head()\n\n# On peut exprimer en proportion...\n\nbase.stats.prop('age').head()\n\nbase.stats.prop('age',bins=[0,18,65,101]).plot.area()\n\nfreqs = base.stats.freq('age',bins=[0,18,80,101])\n\nfreqs.head(10)\n\n# ## Autres variables et sous-ensemble\n\npeduc = base.stats.prop('educ',sub='age>=30')[['none','des','dec','uni']].plot.area()\n\n# On peut aussi rajouter des conditions pour regarder des sous-groupes\n\nbase.stats.prop('married',sub='age>18').plot.area()\n\nbase.stats.prop('nkids',sub='age>=18 & age<=45 & male==False').plot.area()\n\n\n\n\n","repo_name":"CREEi-models/simgen","sub_path":"tests/Statistiques.ipynb","file_name":"Statistiques.ipynb","file_ext":"py","file_size_in_byte":2122,"program_lang":"python","lang":"en","doc_type":"code","stars":0,"dataset":"github-jupyter-script","pt":"41"}
+{"seq_id":"5798158749","text":"# 1. What is the role of try and exception block?\n#\n# Ans:\n#\n# try: try contains blocks of code which might raise a exception or may throw a error.\n#\n# exception: except contain a code which executed the exception if try block got an error or exception raise an error if try block codes not work well. If exception raise in try block then except block will be executed.\n\n# +\n#2. What is the syntax for a basic try-except block?\n\n#Ans:\n\ntry:\n statements\nexcept:\n statements\n\n# +\n#Example\n\ntry:\n a = int(input(\"Enter Integer: \"))\n print(a)\nexcept ValueError:\n print(\"Please enter Integer\")\n# -\n\n# 3.What happens if an exception occurs inside a try block and there is no matching except block?\n#\n# Ans:\n#\n# If an exception occurs inside a try block and there is no matching except block then it will pass error to the outer try statement and it will throw the actual error name. As we can see in this example:\n\ntry:\n a = int(input(\"Enter Integer: \"))\n print(a)\nexcept FileNotFoundError:\n print(\"Please enter Integer\")\n\n# 4. What is the difference between using a bare except block and specifying a specific exception type?\n#\n# Ans:\n#\n# Using bare except block with Exception class, Exception will caught any type of error if we miss some specific type of exception. If we use Exception class then our program will run smoothly without any interruptions.\n#\n# If we use specific exception type and if we miss some exception then program will be interrupted. Because we dont know what type of data a end user will input and also it is not possible to specify all exception in program. So program may get interrupted.\n\n# +\n#5. Can you have nested try-except blocks in Python? If yes, then give an example.\n\n#Example\ntry:\n f = open(\"test.txt\", \"r\") # here we have use r instead of w\n f.write(\"this is my code with exception handling\")\n print(\"this is my code after write ops\")\nexcept Exception as e:\n print(\"here we have use r instead of w: \", e)\ntry:\n print()\n print(\"Program 2\")\n l = [1,2,3,4,5]\n for i in l:\n print(i)\nexcept:\n print(\"this is a handler for loop\")\n\n# +\n# 6. Can we use multiple exception blocks, if yes then give an example.\n\n# Example:\n\ntry:\n a = int(input(\"Enter Integer1: \"))\n b = int(input(\"Enter Integer2: \"))\n print(a, b)\n result = a/b\n print(result)\n \nexcept ValueError:\n print(\"Please enter Integer\")\nexcept ZeroDivisionError:\n print(\"Please don't enter zero\")\n# -\n\n# 7. Write the reason due to which following errors are raised:\n#\n# a. EOFError: End of line Error executed when a end user will not give any input or print function print data with no data given. This error is sometime experienced while using online IDEs.\n\n# +\n# Example of EOFError\n\ntry:\n num = int(input())\n print(num * 10)\nexcept EOFError:\n print(\"no data provided to input function\") # this error occurs in online IDEs\n# -\n\n# b. FloatingPointError: Raised when a floating point operation fails. an error that occurs when a floating-point calculation is attempted that is not supported by the available hardware or software.\n\n# Example of FloatingPointError\n1.2 - 1.0\n\n# c. IndexError: Its occurs when index is out of range\n\n# +\n# Example of IndexError\n\ntry:\n l = [34,56,1,7,8]\n print(l[6])\nexcept IndexError:\n print(\"index out of range\")\n# -\n\n# d. MemoryError: It raise when operations memory runs out of range.\n\n# Example of MemoryError\ntry:\n large_list = [1] * (10 ** 12)\nexcept MemoryError:\n print(\"memory runs out of range.\")\n\n# e. OverflowError: When any operations or arithmetic operation are storing any value above the limit or too large to expressed by the noraml number format.\n#\n\n# +\n# Example of OverflowError\n\ntry:\n resul = 9000 ** 12\nexcept OverflowError:\n print(\"Over flow or above the limit\")\nelse:\n print(resul)\n \n# -\n\n# f. TabError: Raised when indentation contains mixed tabs and spaces.\n\n# g. ValueError: Raised when wrong value is assigned to the variable.\n\n# Example of ValueError\ntry:\n a = int(input(\"Enter Integer: \"))\n print(a)\nexcept ValueError:\n print(\"Please enter Integer\")\n\n# 8. Write code for the following given scenario and add try-exception block to it.\n#\n\n# +\n# a. Program to divide two numbers\n\ntry:\n n1 = int(input(\"Enter number n1: \"))\n n2 = int(input(\"Enter number n2: \"))\n div = n1 / n2\nexcept ValueError:\n print(\"Please enter integer\")\nexcept ZeroDivisionError:\n print(\"Please dont enter zero\")\nelse:\n print(div)\n\n# +\n# b. Program to convert a string to an integer\n\ntry:\n st = input(\"Enter: \")\nexcept ValueError:\n print(\"Please enter string\")\nelse:\n try:\n print(int(st))\n except ValueError:\n print(\"Enter integer to conver sting number into integer\")\n\n# +\n# c. Program to access an element in a list\n\ntry:\n l1 = [12,32,9,45,78]\n print(l1[2])\nexcept IndexError:\n print(\"Index out of range\")\n\n# +\n# d. Program to handle a specific exception\n\ntry:\n f = open(\"test67.txt\", 'r')\n f.write(\"Hello Ineuron\")\nexcept FileNotFoundError:\n print(\"Raised a File not Found Error\")\n\n# +\n# e. Program to handle any exception\n\ntry:\n f = open(\"test67.txt\", 'r')\n f.write(\"Hello Ineuron\")\n\nexcept Exception as e:\n print(\"Raised a File not Found Error: \", e)\n \ntry:\n a = int(input(\"Enter Integer: \"))\n print(a)\nexcept Exception as e:\n print(\"Please enter Integer: \", e)\n \n# -\n\n\n","repo_name":"birkat/Python_Basic_Assignments_2023","sub_path":"Assigment_10_17th_june.ipynb","file_name":"Assigment_10_17th_june.ipynb","file_ext":"py","file_size_in_byte":5401,"program_lang":"python","lang":"en","doc_type":"code","stars":0,"dataset":"github-jupyter-script","pt":"43"}
+{"seq_id":"14228357818","text":"# # Определение перспективного тарифа для телеком-компании\n\n# ## Шаг 1. Обзор данных\n\n# Составление первого представления о данных статистики архива 500 пользователей «Мегалайна»: кто они, откуда, каким тарифом пользуются, сколько звонков и сообщений каждый отправил за 2018 год.\n\n# импорт необходимых библиотек\nimport pandas as pd\nimport numpy as np\nimport matplotlib.pyplot as plt\nimport seaborn as sns\nimport warnings\nfrom scipy import stats as st\n\n\n# функция для просмотра информации и первых строк таблиц\ndef inform(df):\n df.info()\n print('*'*50)\n display(df.head())\n return\n\n\n# Чтение файлов users.csv, calls.csv, messages.csv, internet.csv, tariffs.csv и сохрание их в переменных:\n\n# чтение файлов с данными и сохранение в переменных\nusers_df, calls_df, messages_df, internet_df, tariffs_df = (pd.read_csv('/datasets/users.csv'),\n pd.read_csv('/datasets/calls.csv'),\n pd.read_csv('/datasets/messages.csv'),\n pd.read_csv('/datasets/internet.csv'),\n pd.read_csv('/datasets/tariffs.csv'))\n\n# общая информация и первые 5 строк таблицы users_df\ninform(users_df)\n\n# общая информация и первые 5 строк таблицы calls_df\ninform(calls_df)\n\n# общая информация и первые 5 строк таблицы messages_df\ninform(messages_df)\n\n# общая информация и первые 5 строк таблицы internet_df\ninform(internet_df)\n\n# общая информация и первые 5 строк таблицы tariffs_df\ninform(tariffs_df)\n\n# **Согласно документации к данным:**\n#\n# ***Таблица users_df (информация о пользователях):***\n# * `user_id` — уникальный идентификатор пользователя;\n# * `first_name` — имя пользователя;\n# * `last_name` — фамилия пользователя;\n# * `age` — возраст пользователя (годы);\n# * `reg_date` — дата подключения тарифа (день, месяц, год);\n# * `churn_date` — дата прекращения пользования тарифом (если значение пропущено, то тариф ещё действовал на момент выгрузки данных);\n# * `city` — город проживания пользователя;\n# * `tariff` — название тарифного плана.\n#\n# ***Таблица calls_df (информация о звонках):***\n# * `id` — уникальный номер звонка;\n# * `call_date` — дата звонка;\n# * `duration` — длительность звонка в минутах;\n# * `user_id` — идентификатор пользователя, сделавшего звонок.\n#\n# ***Таблица messages_df (информация о сообщениях):***\n# * `id` — уникальный номер сообщения;\n# * `message_date` — дата сообщения;\n# * `user_id` — идентификатор пользователя, отправившего сообщение.\n#\n# ***Таблица internet_df (информация об интернет-сессиях):***\n# * `id` — уникальный номер сессии;\n# * `mb_used` — объём потраченного за сессию интернет-трафика (в мегабайтах);\n# * `session_date` — дата интернет-сессии;\n# * `user_id` — идентификатор пользователя.\n#\n# ***Таблица tariffs_df (информация о тарифах):***\n# * `tariff_name` — название тарифа;\n# * `rub_monthly_fee` — ежемесячная абонентская плата в рублях;\n# * `minutes_included` — количество минут разговора в месяц, включённых в абонентскую плату;\n# * `messages_included` — количество сообщений в месяц, включённых в абонентскую плату;\n# * `mb_per_month_included` — объём интернет-трафика, включённого в абонентскую плату (в мегабайтах);\n# * `rub_per_minute` — стоимость минуты разговора сверх тарифного пакета (например, если в тарифе 100 минут разговора в месяц, то со 101 минуты будет взиматься плата);\n# * `rub_per_message` — стоимость отправки сообщения сверх тарифного пакета;\n# * `rub_per_gb` — стоимость дополнительного гигабайта интернет-трафика сверх тарифного пакета (1 гигабайт = 1024 мегабайта).\n\n# При просмотре данных в таблицах, можно сделать вывод о том, что в названиях столбцов нарушений нет, тип данных соответсвует их содержимому в столбцах.\n# Для удобства исследования таблицы будут объедины и уже исходя из объединенной таблицы можно будет рассматривать наличие пропусков, дубликатов и ошибок в данных.\n\n# ## Шаг 2. Предобработка данных\n\n# Смена типа данных в столбцах таблиц с датой:\n\n# смена типа данных на DateTime\nusers_df['reg_date'] = pd.to_datetime(users_df['reg_date'], format='%Y.%m.%d %H:%M:%S')\ncalls_df['call_date'] = pd.to_datetime(calls_df['call_date'], format='%Y.%m.%d %H:%M:%S')\nmessages_df['message_date'] = pd.to_datetime(messages_df['message_date'], format='%Y.%m.%d %H:%M:%S')\ninternet_df['session_date'] = pd.to_datetime(internet_df['session_date'], format='%Y.%m.%d %H:%M:%S')\n\n# Выделение месяца для таблиц:\n\ncalls_df['month'] = pd.DatetimeIndex(calls_df['call_date']).month\nmessages_df['month'] = pd.DatetimeIndex(messages_df['message_date']).month\ninternet_df['month'] = pd.DatetimeIndex(internet_df['session_date']).month\n\n# округление длительности звонков\ncalls_df['duration'] = np.ceil(calls_df['duration'])\n\n# +\n# группировка для calls по user_id и month\ncalls = calls_df.groupby(['user_id', 'month']).agg({'duration': ['sum','count']}).reset_index()\ncalls.columns = ['user_id', 'month', 'duration_sum', 'duration_count'] # названия колонок\n\n# группировка для internet по user_id и month\ninternet = internet_df.groupby(['user_id', 'month']).agg({'mb_used': 'sum'}).reset_index()\n\n# группировка для messages по user_id и month и переименование колонки id\nmessages = messages_df.groupby(['user_id', 'month']).agg({'id': 'count'}).rename(columns = {'id': 'messages'}).reset_index()\n# -\n\n# объединение таблиц\nreport = calls.merge(internet, on = ['user_id', 'month'], how = 'outer')\nreport1 = report.merge(messages, on = ['user_id', 'month'], how = 'outer').fillna(0)\nreport2 = report1.merge(users_df, on = 'user_id', how = 'left')\ndf = report2.merge(tariffs_df, left_on = 'tariff', right_on = 'tariff_name', how = 'left')\n\n# общая информация и первые 5 строк таблицы df\ninform(df)\n\n# +\n# перевод длительности звонков в int\ndf['duration_sum'] = df['duration_sum'].astype(int)\n\n# перевод количества звонков в int\ndf['duration_count'] = df['duration_count'].astype(int)\n\n# перевод затраченного интернета в Гб\ndf['mb_used'] = df['mb_used']/1024\n\n# округление затраченного интернета в большую сторону и преведение к целому значению\ndf['mb_used'] = df['mb_used'].apply(np.ceil).astype(int)\n\n# перевод предоставленного интернета в Гб и перевод значения в целочисленное\ndf['mb_per_month_included'] = (df['mb_per_month_included']/1024).astype(int)\n\n# переименование столбцов с интернетом\ndf = df.rename(columns = {\n 'mb_used': 'gb_used', \n 'mb_per_month_included': 'gb_per_month_included'})\n\n# смена типа данных в столбце messages на целочисленный\ndf['messages'] = df['messages'].astype(int)\n\n# проверка\ndisplay(df.head())\n# -\n\n# Посчет для каждого пользователя:\n# * количество сделанных звонков и израсходованных минут разговора по месяцам;\n# * количество отправленных сообщений по месяцам;\n# * объем израсходованного интернет-трафика по месяцам;\n# * помесячную выручку с каждого пользователя (вычтите бесплатный лимит из суммарного количества звонков, сообщений и интернет-трафика; остаток умножьте на значение из тарифного плана; прибавьте абонентскую плату, соответствующую тарифному плану).\n\n# выделение необходимых столбцов в одну таблицу\nreport = df[['user_id', 'last_name', 'month', 'duration_sum', 'duration_count', 'messages', 'gb_used']]\ndisplay(report)\n\n\n# Подсчет помесячной выручки с каждого пользователя:\n\n# создание функции для подсчета помесячной выручки с каждого пользователя\ndef revenue(row):\n \n minute_left = 0\n gd_left = 0\n message_left = 0\n \n if row['duration_sum'] > row['minutes_included']:\n minute_left = (row['duration_sum'] - row['minutes_included']) * row['rub_per_minute']\n \n if row['messages'] > row['messages_included']:\n message_left = (row['messages'] - row['messages_included']) * row['rub_per_message']\n \n if row['gb_used'] > row['gb_per_month_included']:\n gd_left = (row['gb_used'] - row['gb_per_month_included']) * row['rub_per_gb']\n \n \n result = row['rub_monthly_fee'] + minute_left + gd_left + message_left\n \n return result\n\n# создание столбца с помесячной выручкой для каждого пользователя\ndf['revenue_per_month'] = df.apply(revenue, axis = 1)\ndisplay(df)\n\n# расчет срадней выручки по тарифам\nprint(df.query('tariff == \"ultra\"')['revenue_per_month'].mean().round(2))\nprint(df.query('tariff == \"smart\"')['revenue_per_month'].mean().round(2))\n\n# При подготовке данных таблицы с данными по пользователям, сообщениям, звонкам, интернету и тарифам были объединены в одну таблицу, были рассчитаны для каждого пользователя следующие показатели:\n# * количество сделанных звонков и израсходованных минут разговора по месяцам;\n# * количество отправленных сообщений по месяцам;\n# * объем израсходованного интернет-трафика по месяцам;\n# * помесячную выручку с каждого пользователя.\n#\n# Изменен тип данных для столбцов: \n# * `duration`(в целочисленный с округлением в большую сторону), \n# * `mb_used`=`gb_used`(переведен в Гб с округлением в большую сторону и переведен в целочисленный тип), \n# * `messages`(в целочисленный тип),\n# * `reg_date`(в формат DateTime, для выделения месяцев),\n# * `mb_per_month_included`=`gb_per_month_included`(переведен в Гб и преведен в целочисленный тип).\n\n# ## Шаг 3. Анализ данных\n\n# Описание поведения клиентов оператора, исходя из выборки. Сколько минут разговора, сколько сообщений и какой объём интернет-трафика требуется пользователям каждого тарифа в месяц? Посчет среднего количества, дисперсии и стандартного отклонения.\n\n# посчет среднего количества, дисперсии и стандартного отклонения\nmmi_df = df.groupby('tariff_name')[['duration_sum', 'messages', 'gb_used']].agg({'count', 'mean', 'var', 'std'}).style.format('{:.0f}')\ndisplay(mmi_df)\n\n# Выборки по тарифам «Смарт»(\"smart\") и «Ультра»(\"ultra\") (количество значений 2229 и 985, соответственно) по количеству минут разговора `duration_sum`, пользователи тарифа «Смарт» в среднем используют 418, а пользователи тарифа «Ультра» - 527, дисперсия 36219 и 100874, соответственно, это говорит о том, что значения в выборках совсем неодинаковы между собой поэтому их разброс от среднего так велик. \n#\n# По количеству сообщений `messages` пользователи тарифа «Смарт» в среднем используют 33, а пользователи тарифа «Ультра» - 49, дисперсия 797 и 2285, соответственно, вывод о дисперсии такой же, как и в случае по количеству минут разговора.\n#\n# По объёму интернет-трафика `gb_used` пользователи тарифа «Смарт» в среднем используют 16, а пользователи тарифа «Ультра» - 19, дисперсия 33 и 97, соответственно, вывод о дисперсии такой же, как и в случае по количеству минут разговора и по количеству смс.\n\n# Построение гистограммы:\n\n# +\n# для того чтобы предупреждение не мешало\nwarnings.filterwarnings('ignore')\nparams = ['duration_sum', 'messages', 'gb_used']\n\n# число рядов и столбцов в сетке графиков\ncol_count, row_count = 2, 2\n\n# размер графиков\nplt.figure(figsize = (14, 7))\n\n# положение графиков в сетке, i+1 порядковый номер графика\nfor i, param in enumerate(params):\n for tariff in df['tariff_name'].unique():\n current = df.query('tariff_name == @tariff')\n sns.distplot(current[param], label = tariff, \n ax = plt.subplot(row_count, col_count, i + 1)\n )\n \n # добавление подзаголовков на каждый график\n plt.title(param)\n \n # добавление легенды на каждый график\n plt.legend()\n\n# название \nplt.suptitle('Распределение параметров с разбивкой по тарифам')\n\n# настройка подзаголовков\nplt.tight_layout()\n# -\n\n# Описание распределения:\n#\n# На графиках с минутами разговора(`duration_sum`) и объемом интернет-трафика (`gb_used`) распределение значений биномиально со стремлением к нормальному распределению. На графике с количеством сообщений(`messages`) значения распределены биномиально. \n\n# **Вывод**\n#\n# При описание поведения клиентов оператора, исходя из выборки можно сделать вывод о том, что в среднем для клиентов тарифа «Смарт»(\"smart\") хватает минут и сообщений, включенных в тариф, а вот бсплатного интернет-трафика (15 Гб) - недостаточно.\n\n# ## Шаг 4. Проверка гипотез\n\n# *Гипотеза №1: Средняя выручка пользователей тарифов «Ультра» и «Смарт» различаются:*\n#\n# H₀: средняя выручка пользователей тарифа «Ультра» не равна средней выручке пользователей тарифа «Смарт»;\n#\n# H₁: средняя выручка пользователей тарифа «Ультра» равна средней выручке пользователей тарифа «Смарт».\n#\n# Исходя из формулировки гипотезы о том, что средняя выручка пользователей тарифов «Ультра» и «Смарт» различаются, были сформулированы нулевая(H₀) и альтернативная гипотезы(H₁), для нулевой сделано предположение о том, что средняя выручка пользователей тарифа «Ультра» не равна средней выручке пользователей тарифа «Смарт», а для альтернативной, наоборот, средняя выручка пользователей тарифа «Ультра» равна средней выручке пользователей тарифа «Смарт».\n#\n# Генеральные совокупности не зависят друг от друга. Выборочные средние нормально распределены. Для проверки гипотезы подойдет t-тест.\n\n# +\n# критический уровень статистической значимости\n# если p-value окажется меньше него - гипотеза отвергается\nalpha = .001\n\nresults = st.ttest_ind(df.query('tariff_name == \"ultra\"')['revenue_per_month'], \n df.query('tariff_name == \"smart\"')['revenue_per_month']\n )\n\nprint('p-значение: ', results.pvalue)\n\nif results.pvalue < alpha:\n print(\"Отвергаем нулевую гипотезу\")\nelse:\n print(\"Не получилось отвергнуть нулевую гипотезу\") \n# -\n\n# По результату t-теста средняя выручка пользователей тарифа «Ультра» равна средней выручке пользователей тарифа «Смарт». То есть можно сказать, что гипотеза о том, что средняя выручка пользователей тарифов «Ультра» и «Смарт» различаются не подтверждается.\n\n# *Гипотеза№2: Средняя выручка пользователей из Москвы отличается от выручки пользователей из других регионов:*\n#\n# H₀: средняя выручка пользователей из Москвы не равна средней выручке пользователей из других регионов (средние по двум выборкам не равны);\n#\n# H₁: средняя выручка пользователей из Москвы равна средней выручке пользователей из других регионов (средние по двум выборкам равны).\n#\n# Исходя из формулировки гипотезы о том, что средняя выручка пользователей из Москвы отличается от выручки пользователей из других регионов, были сформулированы нулевая(H₀) и альтернативная гипотезы(H₁), для нулевой сделано предположение о том, что средняя выручка пользователей из Москвы не равна средней выручке пользователей из других регионов, а для альтернативной, наоборот, средняя выручка пользователей из Москвы равна средней выручке пользователей из других регионов.\n\n# +\n# критический уровень статистической значимости\n# если p-value окажется меньше него - гипотеза отвергается\nalpha = .001\n\nresults = st.ttest_ind(df.query('city == \"Москва\"')['revenue_per_month'], \n df.query('city != \"Москва\"')['revenue_per_month']\n )\n\nprint('p-значение: ', results.pvalue)\n\nif results.pvalue < alpha:\n print(\"Отвергаем нулевую гипотезу\")\nelse:\n print(\"Не получилось отвергнуть нулевую гипотезу\") \n# -\n\n# По результату t-теста средняя выручка пользователей из Москвы не равна средней выручке пользователей из других регионов. То есть можно сказать, что гипотеза о том, что средняя выручка пользователей из Москвы отличается от выручки пользователей из других регионов не отвергается.\n\n# ## Шаг 5. Общий вывод\n\n# В начале проекта, при просмотре данных в таблицах в названиях столбцов не было обнаружено нарушений стиля, тип данных соответсвовал их содержимому в столбцах. Для удобства исследования таблицы были объедины и уже исходя из объединенной таблицы было рассматрено наличие пропусков, дубликатов и ошибок в данных. При подготовке данных таблицы с данными по пользователям, сообщениям, звонкам, интернету и тарифам были объединены в одну таблицу, были рассчитаны количество сделанных звонков и израсходованных минут разговора, количество отправленных сообщений, объем израсходованного интернет-трафика, помесячная выручка с каждого пользователя по месяцам.\n# Изменены типы данных в столбцах `duration`, `mb_used`(`gb_used`), `messages`, `reg_date`, `mb_per_month_included`(`gb_per_month_included`).\n#\n# При описание поведения клиентов оператора, исходя из выборки, в среднем для клиентов тарифа «Смарт» хватает минут и сообщений, включенных в тариф, а вот бсплатного интернет-трафика (15 Гб) - недостаточно.\n#\n# Исходя из этого, можно сказать, что пользователи более дорогого (по ежемесячной оплате) тарифа «Ультра», в среднем, используют меньше предоставленных им минут, сообщений и интернета, тогда как пользователи тарифа «Смарт», в среднем, используют интернет сверх предоставленного, который оплачивают сверх ежемесячной оплаты по тарифу. При проверке гипотиз, гипотиза о том, что средняя выручка пользователей тарифов «Ультра» и «Смарт» различаются - не подтверждается. В связи с этим можно сказать, что оба тарифа приносят хорошую выручку оператору и при расчете бюджета на рекламу, стоит вложиться в рекламу обоих трифов. Но стоит задуматься о рекламе в регионах, так как гипотиза о том, что средняя выручка пользователей из Москвы отличается от выручки пользователей из других регионов не отвергается.\n#\n","repo_name":"FAnastasiiaD/portfolio","sub_path":"promising_tariff_for_a_telecom_company/promising_tariff_for_a_telecom_company_project.ipynb","file_name":"promising_tariff_for_a_telecom_company_project.ipynb","file_ext":"py","file_size_in_byte":26250,"program_lang":"python","lang":"en","doc_type":"code","stars":0,"dataset":"github-jupyter-script","pt":"43"}
+{"seq_id":"39244058635","text":"import pandas as pd\n\nhd=pd.read_csv(\"E:/datasets/HeartDisease.csv\")\n\nhd.head()\n\nhd.isnull().sum()\n\nfrom sklearn.model_selection import train_test_split\n\nhd_train,hd_test=train_test_split(hd,test_size=0.2,random_state=555)\nhd.shape\n\nhd_train_x=hd_train.iloc[:,0:-1]\nhd_train_y=hd_train.iloc[:,-1]\nhd_test_x=hd_test.iloc[:,0:-1]\nhd_test_y=hd_test.iloc[:,-1]\n\nfrom sklearn.linear_model import LogisticRegression\nlg=LogisticRegression()\n\nlg.fit(hd_train_x,hd_train_y)\n\npred_test=lg.predict(hd_test_x)\n\n# +\nfrom sklearn.metrics import confusion_matrix\n\ntab1 = confusion_matrix(pred_test, hd_test_y)\ntab1\n# -\n\nfrom sklearn.metrics import accuracy_score\naccuracy_score(hd_test_y,pred_test)\n\nfrom sklearn.metrics import precision_score\nprecision_score(hd_test_y,pred_test)\n\nlg.coef_\n\nlg.intercept_\n\nlg.predict_proba(hd_test_x) \n\npred_test\n\n## Build plot AUROC\nfrom sklearn.metrics import roc_auc_score\nfrom sklearn.metrics import roc_curve\n\nlg_auc_roc=roc_auc_score(hd_test_y,pred_test)\nlg_auc_roc\n\nprob=lg.predict_proba(hd_test_x)\n\nprob[:,1]\n\nfpr,tpr,threshold=roc_curve(hd_test_y, prob[:,1])\n\nimport matplotlib.pyplot as plt\nplt.plot(fpr,tpr)\nplt.xlabel(\"False Positive Ratio\",size=20)\nplt.ylabel(\"True Positive Ratio\",size=20)\nplt.title(\"FPR Vs TPR\",size=20)\nplt.grid()\n\n\n","repo_name":"MrunalineePatole/Logistic","sub_path":"Logistic_Heart_Disease.ipynb","file_name":"Logistic_Heart_Disease.ipynb","file_ext":"py","file_size_in_byte":1269,"program_lang":"python","lang":"en","doc_type":"code","stars":0,"dataset":"github-jupyter-script","pt":"43"}
+{"seq_id":"3466337714","text":"# +\n#Data Manipulation libraries \nimport numpy as np\nimport pandas as pd\n\n#Data Visualisation libraries\nimport matplotlib.pyplot as plt\nfrom matplotlib import rcParams\nimport seaborn as sns \n\n#Applying cool styles \nplt.style.use(\"ggplot\")\nrcParams['figure.figsize']=(12, 6)\n\n\n# from sklearn.model_selection import train_test_split\n# from sklearn import linear_model\nfrom sklearn.cluster import KMeans\nfrom sklearn.preprocessing import MinMaxScaler\n# -\n\ndf = pd.read_csv(\"datasets/animes.csv\")\ndf.head(3)\n\nepisodes = df[\"episodes\"].values\ndf[\"episodes\"] = df[\"episodes\"].fillna(np.mean(df[\"episodes\"]))\n\n# +\nbeginner = []\npro = []\nadvanced = []\nl_distances = []\n\nfor i in episodes:\n if i <= 100:\n l_distances.append(((((100-0)/2)-i)**2)**0.5) \n # 50 as centroid for anime with eps lessthan 100\n elif i <= 500:\n l_distances.append(((((500-0)/2)-i)**2)**0.5) \n # 250 as centroid for anime with eps lessthan 100\n else:\n l_distances.append(((((3057-0)/2)-i)**2)**0.5)\n # 3057(max eps for an anime)//2 as centroid for rest\n\nfor i in l_distances:\n print(i)\n# -\n\ndf[\"Distance\"] = l_distances\ndf[['Distance']] = df[['Distance']].apply(pd.to_numeric) \ndf[\"Distance\"] = df[\"Distance\"].fillna(np.mean(df[\"Distance\"]))\n\ndf.head(3)\n\nplt.scatter(df['episodes'],df['Distance'])\n\nkm = KMeans(n_clusters = 3)\nkm\n\ny_predicted = km.fit_predict(df[[\"episodes\",'Distance']])\ny_predicted\n\ndf[\"cluster\"] = y_predicted\ndf.head(3)\n\nkm.cluster_centers_\n\ndf1 = df[df.cluster==0]\ndf2 = df[df.cluster==1]\ndf3 = df[df.cluster==2]\nplt.scatter(df1.episodes,df1['Distance'],color='green')\nplt.scatter(df2.episodes,df2['Distance'],color='red')\nplt.scatter(df3.episodes,df3['Distance'],color='black')\nplt.scatter(km.cluster_centers_[:,0],km.cluster_centers_[:,1],color='purple',marker='*',label='centroid')\nplt.xlabel('episodes')\nplt.ylabel('Distance')\nplt.legend()\n\n# ## Preproccessing using min and max scaler\n\n# +\nscaler = MinMaxScaler()\n\nscaler.fit(df[['Distance']])\ndf['Distance'] = scaler.transform(df[['Distance']])\n\nscaler.fit(df[['episodes']])\ndf['episodes'] = scaler.transform(df[['episodes']])\n# -\n\nplt.scatter(df['episodes'],df['Distance'])\n\n\n\n\n\n\n\n\n\n\n","repo_name":"ramkiran55/Anime_CRP","sub_path":"DataMiningClustering.ipynb","file_name":"DataMiningClustering.ipynb","file_ext":"py","file_size_in_byte":2181,"program_lang":"python","lang":"en","doc_type":"code","stars":0,"dataset":"github-jupyter-script","pt":"43"}
+{"seq_id":"715904178","text":"# ### Yielding and Generators\n\n# Let's start by writing a \"simple\" iterator first using the techniques we learned in the previous section.\n\nimport math\n\n\nclass FactIter:\n def __init__(self, n):\n self.n = n\n self.i = 0\n\n def __iter__(self):\n return self\n\n def __next__(self):\n if self.i >= self.n:\n raise StopIteration\n else:\n result = math.factorial(self.i)\n self.i += 1\n return result\n\n\nfact_iter = FactIter(5)\n\nfor num in fact_iter:\n print(num)\n\n\n# We could achieve the same thing using the `iter` method's second form - we just have to know our sentinel value - in this case it would be the factorial of n+1 where n is the last integer's factorial we want our iterator to produce:\n\ndef fact():\n i = 0\n def inner():\n nonlocal i\n result = math.factorial(i)\n i += 1\n return result\n return inner \n\n\nfact_iter = iter(fact(), math.factorial(5))\n\nfor num in fact_iter:\n print(num)\n\n\n# You'll note that in both cases `fact_iter` was an **iterator**. In the first example we implemented the iterator ourselves, in the second example Python built-it for us.\n#\n# The second example was a little less code, but maybe a little more difficult to understand if we were just shown the code without having written it ourselves.\n#\n# There has to be a better way!!\n\n# And indeed, there is... generators.\n\n# Let's look at the `yield` statement first.\n#\n# The `yield` statement is used almost like a `return` statement in a function - but there is a huge difference - when the `yield` statement is encountered, Python returns whatever value `yield` specifies, but it \"pauses\" execution of the function. We can then \"call\" the same function again and it will \"resume\" from where the last `yield` was encountered.\n#\n# I say \"call\" because we do not \"resume\" the function by calling it - instead we use the function... `next()` !!!\n#\n# Let's try it:\n\ndef my_func():\n print('line 1')\n yield 'Flying'\n print('line 2')\n yield 'Circus' \n\n\nmy_func()\n\n# So, executing `my_func()`, returned a generator object - it did not actually \"run\" the body of `my_func` (none of our print statements actually ran).\n#\n# To do that, we need to use the `next()` function. \n#\n# `next()`?? Isn't that what we use for iteration??\n\ngen_my_func = my_func()\n\nnext(gen_my_func)\n\nnext(gen_my_func)\n\n# And let's call it one more time:\n\nnext(gen_my_func)\n\n# A `StopIteration` exception.\n#\n# Hmmm... `next`, `StopIteration`? What does this look like? \n#\n# An **iterator**!\n\n# And in fact that's exactly what Python generators are - they **are** iterators. \n\n# If generators are iterators, they should implement the iterator **protocol**.\n#\n# Let's see:\n\ngen_my_func = my_func()\n\n'__iter__' in dir(gen_my_func)\n\n'__next__' in dir(gen_my_func)\n\n# And so we just have an iterator, which we can use with the `iter()` function and the `next()` function like any other iterator:\n\ngen_my_func\n\niter(gen_my_func)\n\n\n# As you can see, the `iter` function returned the same object - something we expect with iterators.\n\n# So if this is an iterator that Python builds, how does it know when to stop the iteration (raise the `StopIteration` exception)?\n#\n# In the example above, it seemed clear - when the function finished running - there were no more statements after that last `yield`.\n#\n# What actually happens if a function finishes running and we don't explicitly return something?\n#\n# Remember that Python fills in the gap, and returns `None`.\n#\n# In general, the iteration will terminate when we **return** something from the function.\n#\n# Let's take a look:\n\ndef squares(sentinel):\n i = 0\n while True:\n if i < sentinel:\n result = i**2\n i += 1\n yield result\n else:\n return 'all done!'\n\n\nsq = squares(3)\n\nnext(sq)\n\nnext(sq)\n\nnext(sq)\n\nnext(sq)\n\n\n# And the return value of our function became the message of the `StopIteration` exception.\n\n# But, we can simplify this slightly:\n\ndef squares(sentinel):\n i = 0\n while True:\n if i < sentinel:\n yield i**2\n i += 1 # note how we can incremenet **after** the yield\n else:\n return 'all done!'\n\n\nfor num in squares(5):\n print(num)\n\n\n# So now let's see how we could re-write our initial `factorial` example:\n\ndef factorials(n):\n for i in range(n):\n yield math.factorial(i) \n\n\nfor num in factorials(5):\n print(num)\n\n# Now that's a much simpler and understandable way to create the iterator!\n\n# Note that a generator **is** an iterator, but not vice-versa - iterators are not necessarily generators, just like sequences are iterables, but iterables are not necessarily sequences.\n\n# Another thing to note is that since generators are iterators, they also become exhausted (consumed) just like an iterator does.\n\nfacts = factorials(5)\n\nlist(facts)\n\nlist(facts)\n\n# As you can see, our second iteration through the same generator ended up with nothing - that's because the generator has been exhausted:\n\nnext(facts)\n\n\n","repo_name":"TarrySingh/Artificial-Intelligence-Deep-Learning-Machine-Learning-Tutorials","sub_path":"python-tuts/1-intermediate/03 - Generators/01 - Yielding and Generators.ipynb","file_name":"01 - Yielding and Generators.ipynb","file_ext":"py","file_size_in_byte":5088,"program_lang":"python","lang":"en","doc_type":"code","stars":3543,"dataset":"github-jupyter-script","pt":"43"}
+{"seq_id":"19487501589","text":"import pandas as pd\nimport numpy as np\nimport requests\nimport math\n\n# ### People\n\nresponse = requests.get('https://swapi.dev/api/')\nlink = response.json()\n\nlink\n\n# +\n#people_df.drop(people_df.index, axis=0, inplace=True)\n# -\n\nresponse = requests.get(link['people'])\npeople = response.json()\n\nnumber_of_people = people['count']\nnext_page = people['next']\nprevious_page = people['previous']\nnumber_of_results = len(people['results'])\nmax_page = math.ceil(number_of_people / number_of_results)\n\npeople_df = pd.DataFrame(people['results'])\n\nfor i in range(max_page-1):\n response = requests.get(people['next'])\n people = response.json()\n people_df = pd.concat([people_df, pd.DataFrame(people['results'])], ignore_index=True)\n\npeople_df.tail()\n\n# ### Planets\n\nresponse = requests.get(link['planets'])\nplanets = response.json()\n\nplanets\n\nnumber_of_planets = planets['count']\n#next_page = planets['next']\n#previous_page = planets['previous']\nnumber_of_results = len(planets['results'])\nmax_page = math.ceil(number_of_planets / number_of_results) #6\n\nplanets_df = pd.DataFrame(planets['results'])\n\nfor i in range(max_page-1):\n response = requests.get(planets['next'])\n planets = response.json()\n planets_df = pd.concat([planets_df, pd.DataFrame(planets['results'])], ignore_index=True)\n\n# +\n#planets_df.index = planets_df.index.factorize()[0]\n# -\n\nplanets_df.tail(2)\n\n# #### Starships\n\nlink\n\nresponse = requests.get(link['starships'])\nstarships = response.json()\n\nnumber_of_starships = starships['count']\n#next_page = planets['next']\n#previous_page = planets['previous']\nnumber_of_results = len(starships['results'])\nmax_page = math.ceil(number_of_starships / number_of_results) #6\n\nmax_page\n\nstarships_df = pd.DataFrame(starships['results'])\n\nfor i in range(max_page-1):\n response = requests.get(starships['next'])\n starships = response.json()\n starships_df = pd.concat([starships_df, pd.DataFrame(starships['results'])], ignore_index=True)\n\nstarships_df.tail(2)\n","repo_name":"nadia-paz/time-series-exercises","sub_path":"api_exercises.ipynb","file_name":"api_exercises.ipynb","file_ext":"py","file_size_in_byte":1983,"program_lang":"python","lang":"en","doc_type":"code","stars":0,"dataset":"github-jupyter-script","pt":"43"}
+{"seq_id":"16827508535","text":"import numpy as np\nimport pandas as pd\nimport matplotlib.pyplot as plt\nimport seaborn as sns\n\ntitanic=pd.read_csv(\"train.csv\")\ntitanic.head()\n\n# ## Analyze the data\n\ntitanic.isna().sum()\n\ntitanic.describe()\n\nsns.countplot(x=\"Sex\", data=titanic, palette=\"RdBu_r\")\n\nsns.countplot(x=\"Survived\", data=titanic, palette=\"RdBu_r\")\n\nsns.countplot(x=\"Sex\",hue=\"Survived\", data=titanic,palette=\"RdBu_r\" )\n\nsns.countplot(x=\"Pclass\", hue=\"Survived\",data=titanic, palette=\"RdBu_r\")\n\ntitanic.corr()\n\nsns.heatmap(titanic.corr(), cmap=\"YlGnBu\")\n\nplt.figure(figsize=(12,6))\nplt.hist(x=\"Age\", data=titanic, bins=25, color=\"r\", alpha=0.5);\n\nsns.countplot(x=\"SibSp\", data=titanic)\n\n# ## Data cleaning\n\n# **Since Age has roughly 20% of missing values, we will replace the missing values. Let's look at how to replace the missing values of the age.**\n\ntitanic.head()\n\nplt.figure(figsize=(12,6))\nsns.boxplot(x=\"Pclass\", y=\"Age\", data=titanic)\n\ntitanic.groupby(by=\"Pclass\").mean(\"Age\")\n\n\n# **Looking at the data above, let's replace the age based on the average of the pclass.**\n\n# +\ndef age_replacement(cols):\n Age=cols[0]\n Pclass=cols[1]\n \n if pd.isnull(Age):\n if Pclass==1:\n return 38\n elif Pclass==2:\n return 30\n else:\n return 25\n \n else:\n return Age\n \n \n# -\n\ntitanic[\"Age\"]=titanic[[\"Age\", \"Pclass\"]].apply(age_replacement, axis=1)\n\ntitanic.isnull().sum()\n\n# **Let's get rid of the cabin column, since there are too many null values.**\n\ntitanic.drop([\"Cabin\"], axis=1,inplace=True)\n\ntitanic.head()\n\ntitanic.dropna(inplace=True)\n\n\n\n# ## Let's get convert categorical features into dummy features using pandas.\n\nsex=pd.get_dummies(titanic[\"Sex\"], drop_first=True)\nembarked=pd.get_dummies(titanic[\"Embarked\"], drop_first=True)\n\ntitanic.drop([\"Sex\", \"Name\", \"Ticket\", \"Embarked\"], axis=1, inplace=True)\n\ntitanic=pd.concat([titanic, sex, embarked], axis=1)\n\ntitanic.head()\n\n# ## Using Logistic Regression\n\nfrom sklearn.model_selection import train_test_split\nfrom sklearn.metrics import classification_report, confusion_matrix, plot_confusion_matrix\nfrom sklearn.linear_model import LogisticRegression\n\n# +\nX=titanic.drop([\"Survived\"],axis=1)\ny=titanic[\"Survived\"]\n\nX_train, X_test, y_train, y_test= train_test_split(X, y, test_size=0.3)\nmodel=LogisticRegression(max_iter=1000)\nmodel.fit(X, y)\npredictions=model.predict(X)\n\nprint(classification_report(y, predictions))\n#plot_confusion_matrix( predictions,X, y, normalize=True)\nprint(confusion_matrix(y, predictions, normalize=\"true\"))\n# -\n\n\n\n\n# # Using NNs\n\nimport tensorflow as tf\nfrom tensorflow.keras.models import Sequential\nfrom tensorflow.keras.layers import Dense, Activation,Dropout\n\nX_train.shape\ntf.keras.backend.set_floatx('float64')\n\n# +\nmodel=Sequential()\n\nmodel.add(Dense(units=20, activation=\"sigmoid\"))\nmodel.add(Dense(units=15,activation='sigmoid'))\nmodel.add(Dense(units=1,activation='sigmoid'))\n\nmodel.compile(loss=\"binary_crossentropy\", optimizer=\"adam\")\n\n# +\nmodel.fit(X_train, \n y=y_train,\n epochs=60,\n validation_data=(X_test, y_test),\n verbose=1\n\n)\n# -\n\nmodel_loss = pd.DataFrame(model.history.history)\n\nmodel_loss.plot()\n\npredictions = model.predict_classes(X_test)\n\nprint(classification_report(y_test, predictions))\nprint(confusion_matrix(y_test, predictions, normalize=\"true\"))\n\n\n\n# ## SVM\n\nfrom sklearn.svm import SVC\n\n\n# +\n\nmodel=SVC()\nmodel.fit(X_train, y_train)\nsvm_predictions = model.predict(X_test)\nprint(classification_report(y_test, svm_predictions))\nprint(confusion_matrix(y_test, svm_predictions, normalize=\"true\"))\n# -\n\n\n\n# ## Catboost\n\nfrom catboost import CatBoostClassifier\n\n# +\nmodel=CatBoostClassifier(verbose=False )\nmodel.fit(X_train, y_train)\ncat_predictions=model.predict(X_test)\n\nprint(classification_report(y_test, cat_predictions))\nprint(confusion_matrix(y_test, cat_predictions, normalize=\"true\"))\n# -\n\n\n","repo_name":"Katerinakz/Python-projects-EDA-ML","sub_path":"Titanic analysis.ipynb","file_name":"Titanic analysis.ipynb","file_ext":"py","file_size_in_byte":3907,"program_lang":"python","lang":"en","doc_type":"code","stars":0,"dataset":"github-jupyter-script","pt":"43"}
+{"seq_id":"4940602888","text":"import pandas as pd\nfrom prophet import Prophet\nfrom statsmodels.tools.eval_measures import rmse, meanabs\n\ndf = pd.read_csv('date_transaction.csv')\ndf['date'] = pd.to_datetime(df['date'])\ndf = df.groupby(pd.Grouper(key='date', freq=\"Y\")).sum().reset_index()\ndf.columns = ['ds', 'y']\ndf\n\ntrain = df[df['ds'] < pd.Timestamp('2021-01-01')]\ntest = df[df['ds'] >= pd.Timestamp('2021-01-01')]\n\nmodel = Prophet(weekly_seasonality=False, changepoint_prior_scale=0.001)\nmodel.fit(train)\n\nfuture = model.make_future_dataframe(3, 'Y')\nfuture.tail()\n\nforecast = model.predict(future)\nfig1 = model.plot(forecast)\nfig2 = model.plot_components(forecast)\n\nforecast[['ds', 'yhat']].tail(5)\n\n# +\npredictions = model.predict(test[['ds']])['yhat']\nactuals = test['y']\n\nprint(f\"Root Mean Squared Error (RMSE): {round(rmse(predictions, actuals))}\")\nprint(f\"Mean Absolute Error (MAE): {round(meanabs(predictions, actuals))}\")\n","repo_name":"isa96/etl-project","sub_path":"results/transaction_forecast.ipynb","file_name":"transaction_forecast.ipynb","file_ext":"py","file_size_in_byte":903,"program_lang":"python","lang":"en","doc_type":"code","stars":1,"dataset":"github-jupyter-script","pt":"43"}
+{"seq_id":"72077619329","text":"import os, glob, numpy as np, pandas as pd\nimport matplotlib.pyplot as plt\nimport time\nplt.rcParams['figure.figsize'] = [25, 15]\nplt.rcParams.update({'font.size': 20})\n\n\ndef load_dataframe_from_files(dirin, fileprefix, max_files=1000):\n import glob\n files = glob.glob(os.path.join(dirin, fileprefix))\n print(\"[Info] Loading {} files wt prefix:\\n{}\".format(len(files), fileprefix))\n df = pd.read_csv(files[0], comment='#', index_col=False)\n for file in files[1:max_files]:\n print(\".\", end='')\n dftmp = pd.read_csv(file, comment='#', index_col=False)\n df = pd.concat([df, dftmp])\n print(\"\")\n return df\n\n\n# +\ninit = time.time()\ndirin = os.path.join(\"..\", \"..\", \"Data\", \"OutputProcessing\", \"Neutrons_VariousConfig_08_20_2020\", \"Neutrons_Sliced\", \"Reduced5MSteps\")\n\nfileinprefix = \"SlicedDetections*100mmCylinder_01MeV*SHORT5MSTEPS.csv\"\ndf_ar41_100cylinder_01Mev = load_dataframe_from_files(dirin, fileinprefix)\ndf_ar41_100cylinder_01Mev = df_ar41_100cylinder_01Mev[df_ar41_100cylinder_01Mev.columns[:-1]]\n\nprint(\"[Info] Loaded data in {:.3f} seconds\".format(time.time() - init))\n# -\n\n# select a random event: eventnumber==87\nevent = df_ar41_100cylinder_01Mev[df_ar41_100cylinder_01Mev.eventnumber==87]\n\nnzero_deteff_steps = event[event.detectionefficiency>0]\n\nbins = np.linspace(0, 2000, 100)\nplt.hist((event.x**2 + event.y**2 + event.z**2)**.5, bins=bins, label=\"Det.Eff.=0\")\nplt.hist((nzero_deteff_steps.x**2 + nzero_deteff_steps.y**2 + nzero_deteff_steps.z**2)**.5, bins=bins, label=\"Det.Eff.>0\")\nplt.legend()\nplt.show()\n\nevent[event.PID==2112]\n\ndf_ar41_100cylinder_01Mev.columns\n\n# # Looking for event\n\ndf = df_ar41_100cylinder_01Mev\n\ndf.columns\n\ndf.tracknumber.unique()\n\n\n","repo_name":"luigiberducci/ML4NP","sub_path":"Notebooks/AnalysisAr41FromNeutronSimulation/reduced_analysis_ar41_for_running_on_laptop.ipynb","file_name":"reduced_analysis_ar41_for_running_on_laptop.ipynb","file_ext":"py","file_size_in_byte":1715,"program_lang":"python","lang":"en","doc_type":"code","stars":0,"dataset":"github-jupyter-script","pt":"43"}
+{"seq_id":"6852122393","text":"import cv2\nfrom matplotlib import pyplot as plt\nimport ov_coefs\nfrom typing import Tuple\nimport numpy\n\n# # Input Data\n\ninput_img = cv2.imread(\"../assets/Van_Gogh_256x256.png\", cv2.IMREAD_COLOR)\ninput_img = cv2.cvtColor(input_img,cv2.COLOR_BGR2RGB)\n# input_img = cv2.resize(input_img, (256, 311))\n\nplt.imshow(input_img)\nplt.show()\nplt.imsave(\"input.png\", input_img)\nprint(input_img.shape)\n\ngray_input_img = cv2.cvtColor(input_img, cv2.COLOR_BGR2GRAY)\nplt.imshow(gray_input_img, \"gray\")\nplt.show()\ncv2.imwrite(\"output_gray.png\", gray_input_img)\nprint(gray_input_img.shape)\n\n\ndef center_crop(img: numpy.ndarray, dim: Tuple[int, int]):\n \"\"\"Returns center cropped image\n Args:\n img: image to be center cropped\n dim: dimensions (width, height) to be cropped\n \"\"\"\n\n width, height = img.shape[1], img.shape[0]\n\n # process crop width and height for max available dimension\n crop_width = dim[0] if dim[0]![](\"../img/image.png\")
\n\n# # Instructions: \n# We are trying to predict who will claim insurance as a function of age using the data. To do so we need :\n# - Read the `insurance_claim.csv` as a dataframe.\n# - Assign the predictor and response variables.\n# - Guesstimate the values of the coefficients $\\hat{\\beta}_0$ and $\\hat{\\beta}_1$.\n# - Predict the response variable using the formula of a simple logistic regression given below (no package allowed) \n# - Compute the accuracy of the model.\n# - Repeat the above steps by changing the values of the coefficients $\\hat{\\beta}_0$ and $\\hat{\\beta}_1$, until you get \"good\" accuracy.\n# - Plot the `Insurance Claim` vs. `Age` graph with the fit of the model.\n#\n# # Hints: \n# - Logistic Regression equation: \n#\n# $$P\\left(Y=1\\right)=\\frac{1}{1+e^{-\\left(\\beta_{0\\ }+\\beta_1X\\right)}}$$\n#\n# plt.plot() : Plots x versus y as lines and/or markers\n#\n# sklearn.accuracy_score() : Accuracy classification score\n#\n# plt.xticks() : Get or set the current tick locations and labels of the x-axis.\n#\n# plt.yticks() : Get or set the current tick locations and labels of the y-axis.\n#\n# plt.xlabel() : Set the label for the x-axis.\n#\n# plt.ylabel() : Set the label for the y-axis.\n#\n# **Note: This exercise is auto-graded and you can try multiple attempts.**\n\n# +\n# import important libraries\n\n# %matplotlib inline\nimport numpy as np\nimport pandas as pd\nfrom math import exp\nimport matplotlib.pyplot as plt\nfrom sklearn.preprocessing import normalize\nfrom sklearn.metrics import accuracy_score\n\n# +\n# Make a dataframe of the file \"insurance_claim.csv\"\n\ndata_filename = 'insurance_claim.csv'\ndf = pd.read_csv(data_filename)\n\n# +\n# Take a quick look of the data, notice that the response variable is binary\n\ndf.head()\n\n# +\n# Assign age as the predictor variable using double brackets\nx = df[[___]]\n\n# Assign insuranceclaim as the response variable\ny = df[___]\n\n\n# +\n# Make a plot of the response (insuranceclaim) vs the predictor (age)\nplt.plot(___,___,'o', markersize=7,color=\"#011DAD\",label=\"Data\")\n\n# Also add the labels for 'x' & 'y' values\nplt.xlabel(___)\nplt.ylabel(___)\n\nplt.xticks(np.arange(18, 80, 4.0))\n\n# Label the value 1 as 'Yes' & 0 as 'No'\nplt.yticks((0,1), labels=('No', 'Yes'))\nplt.legend(loc='best')\nplt.show()\n\n# +\n### edTest(test_beta_guesstimate) ###\n# Guesstimate the values of beta0 & beta1\n\nbeta0 = ___\n\nbeta1 = ___\n\n\n# +\n### edTest(test_beta_computation) ###\n# Use the logistic function below to predict the response based on the input\nlogit = []\n\nfor i in x:\n # Append the P(y=1) values to the logit list\n logit.append(___)\n \n \n\n# +\n# If the predictions are above a threshold of 0.5, predict as 1, else 0\n\ny_pred = []\n\nfor py in ___:\n if py >= 0.5:\n ____\n else:\n ____\n \n\n\n# +\n# Use accuracy_score function to find the accuracy \n\naccuracy = accuracy_score(___, ___)\n\n# Print the accuracy\nprint (___)\n\n# +\n# Make a plot similar to the one above along with the fit curve\nplt.plot(___, ___,'o', markersize=7,color=\"#011DAD\",label=\"Data\")\n\nplt.plot(___,___,linewidth=2,color='black',label=\"Prediction\")\n\nplt.xticks(np.arange(18, 80, 4.0))\nplt.xlabel(\"Age\")\nplt.ylabel(\"Insurance claim\")\nplt.yticks((0,1), labels=('No', 'Yes'))\nplt.legend()\nplt.show()\n# -\n\n# ## Post exercise question:\n#\n# In this exercise, you may have had to stumble around to find the right values of $\\beta_0$ and $\\beta_1$ to get accurate results.\n#\n# Although you may have used visual inspection to find a good fit, in most problems you would need a quantative method to measure the performance of your model. (*Loss function*)\n#\n# Which of the following below are **NOT** possible ways of quantifying the performance of the model.\n#\n# - A. Compute the mean squared error loss of the predicted labels.\n# - B. Evaluate the log-likelihood for this Bernoulli response variable.\n# - C. Go the the temple of Apollo at Delphi, and ask the high priestess Pythia\n# - D. Compute the total number of misclassified labels.\n\n# +\n### edTest(test_quiz) ###\n\n# Put down your answers in a string format below (using quotes)\n\n# for. eg, if you think the options are 'A' & 'B', input below as \"A,B\"\n\nanswer = ___\n","repo_name":"Harvard-IACS/2020-CS109A","sub_path":"docs/lectures/lecture15/notebook/lec15-ex1-challenge.ipynb","file_name":"lec15-ex1-challenge.ipynb","file_ext":"py","file_size_in_byte":5074,"program_lang":"python","lang":"en","doc_type":"code","stars":104,"dataset":"github-jupyter-script","pt":"44"}
+{"seq_id":"22821875045","text":"import pandas as pd\n# hierarchical indices and columns\nindex = pd.MultiIndex.from_product([[2013, 2014], [1, 2]],\n names=['year', 'visit'])\ncolumns = pd.MultiIndex.from_product([['Alice', 'Bob', 'Sue'], ['HR', 'Temp']],\n names=['subject', 'type'])\n\n# mock some data\ndata = np.round(np.random.randn(4, 6), 1)\ndata[:, ::2] *= 10\ndata += 37\n\n# create the DataFrame\nhealth_data = pd.DataFrame(data, index=index, columns=columns)\nhealth_data\n# -\n\n# ## Mean on year rows\n\ndata_mean = health_data.mean(level='year')\ndata_mean\n\n# ## Mean on type columns\n\ndata_mean.mean(axis=1, level='type')\n","repo_name":"jdelano18/cocode","sub_path":"finalbricks/Aggregate Data on Multi-Indices.ipynb","file_name":"Aggregate Data on Multi-Indices.ipynb","file_ext":"py","file_size_in_byte":654,"program_lang":"python","lang":"en","doc_type":"code","stars":1,"dataset":"github-jupyter-script","pt":"44"}
+{"seq_id":"37002312262","text":"import numpy as np\nimport pandas as pd\n\n# Question 1 :\n# Soit la matrice A et la matrice B générées avec le code ci-dessous:\n#\n# A=np.random.RandomState(seed=3).randint(1,100,size=(50,40))\n#\n# B=np.random.RandomState(seed=4).randint(1,100,size=(50,))\n#\n# Additionnez le vecteur B à chaque colonne de la matrice A et affectez le résultat à la matrice C. \n#\n# Inscrire la somme de la première ligne de la matrice C.\n\n# +\nA = np.random.RandomState(seed = 4).randint(10, 41, size = (50, 40))\nB = np.random.RandomState(seed = 4).randint(10, 41, size = (50,))\n\n#print(\"A\", A)\n#print(\"B\", B)\n\nC = A + B[:, np.newaxis] #Addition de B a chaque colonne de la matrice A grâce a l'utilisation de newaxis qui ajoute une nouvelle dimension a ma matrice, ici transformation en colonne\n#et affectation le résultat à la matrice C\n\n#print(\"Bnew\", B[:, np.newaxis])\nprint(np.sum(C[0, :])) #Affichage de la somme de la première ligne de la matrice C\n# -\n\n# Question 2:\n# Soit la matrice A générée avec le code ci-dessous:\n#\n# A=np.random.RandomState(seed=3).randint(1,100,size=(25,50))\n#\n#\n# Inscrire le nombre d'éléments que contient la matrice A.\n\n# +\nA = np.random.RandomState(seed = 3).randint(1, 100, size = (25, 50))\n#print(\"A\", A)\n\nprint(np.size(A)) #affichage de la taille de la matrice A grâce size qui retourne la taille totale du tableau (nombre d’éléments)\n# -\n\n# Question 3 :\n# Soit la matrice A générée avec le code ci-dessous:\n#\n# A=np.random.RandomState(seed=3).randint(1,100,size=(50,40))\n#\n# Calculez la somme des éléments dans la troisième ligne de la matrice A qui sont divisibles par 2 (sans reste).\n\n# +\nA = np.random.RandomState(seed = 3).randint(1, 100, size = (50, 40))\nthird_line = A[2,:] #initialisation de third line en donnant la valeur de la 3eme ligne de A\n#print(\"A\", A)\n#print(\"td\", third_line)\n\nsum = np.sum(third_line[third_line % 2 == 0]) #calcul a l'aide de sum des valeurs divisible par 2\nprint(sum)\n# -\n\n# Question 4:\n# Soit la matrice A générée avec le code ci-dessous: A=np.random.RandomState(seed=3).randint(1,100,size=(50,40)) \n#\n# Ajoutez la somme des éléments de la matrice A compris entre 10 et 20 inclusivement aux éléments de la 4e ligne de cette même matrice. \n#\n# Suite à cette opération, calculez la moyenne de tous les éléments de la matrice A arrondie à la première décimale.\n\n# +\nA = np.random.RandomState(seed = 3).randint(1, 100,size = (50, 40)) \n#print(\"A\", A)\n\nsum = np.sum(A[(A >= 10) & (A <= 20)]) #ajout avec sum des éléments compris entre 10 et 20 donc >= 10 et <= 20\n#print(\"sum\", sum)\n#print(\"third line\",(A[3, :]))\n\n(A[3, :]) = sum + (A[3, :]) #ajout de la somme aux élément de la 4 eme ligne de A\n#print(\"third line\",(A[3, :]))\n\nprint(round(np.mean(A), 1)) #arrondie a la premiere décimale grace a round calcul de la moyenne grace a mean\n# -\n\n# Question 5 :\n# Soit la matrice A et la liste B générées avec le code ci-dessous:\n#\n# A=np.random.RandomState(seed=4).randint(1,10,size=(50,40))\n#\n# B= [1,2,3,4]\n#\n# Sélectionnez parmi les listes suivantes celle qui contient pour chaque élément de B combien de fois ceux-ci se retrouvent dans la matrice A\n\n# +\nA = np.random.RandomState(seed = 4).randint(1, 10, size = (50, 40))\nB = [1,2,3,4]\n\n#print(\"A\", A)\n\n#compte du nombre d'occurrences de la valeur 1 2 3 4 dans la matrice A\nprint(np.count_nonzero(A == 1) , np.count_nonzero(A == 2) , np.count_nonzero(A == 3) , np.count_nonzero(A == 4))\n\n# -\n\n# Question 6:\n# Soit la matrice A générée avec le code ci-dessous:\n#\n# A=np.random.RandomState(seed=4).randint(1,30,size=(50,40))\n#\n# Sélectionnez parmi les choix suivants ceux qui correspondent aux coordonnées (indice de ligne, indice de colonne) de la 5e occurrence de nombre 9 dans la matrice A.\n#\n#\n\n# +\nA = np.random.RandomState(seed = 4).randint(1, 30, size = (50, 40))\n\n#print(\"A\", A)\n\nindices = np.where(A == 9) # utilisation de where en donnant 9 pour avoir un tuple avec 2 tableau d'indice des occurence de 9\n#print(\"Indice\", indices_9)\nrow = indices[0][4] #selection du tableau 1 qui contient les indices de lignes je prends la 5 eme occurence\ncol = indices[1][4] #selection du tableau 2 qui contient les indices de colonne je prends la 5eme occurence\nprint(row, col)\n\n# -\n\n# Question 7 : Utilisez le fichier lesions.xlsx et cause.csv pour répondre à la question.\n#\n# Imputez les valeurs manquantes pour la variable MONTANT_INDEMNITÉ_VERSÉ par la moyenne de celle-ci dans l'industrie (variable INDUSTRIE) de l'observation pour laquelle la valeur est manquante. Affectez le résultat de cette opération à la variable montant_indemnite_modif. \n#\n# Calculez la moyenne de cette nouvelle variable (montant_indemnite_modif) arrondie à la deuxième décimale.\n#\n# Attention: vous ne devez pas modifier les valeurs de la colonne MONTANT_INDEMNITÉ_VERSÉ\n\n# +\nlesion = pd.read_excel(\"lesions.xlsx\")\n\nlesion['montant_indemnite_modif'] = lesion['MONTANT_INDEMNITÉ_VERSÉ'] #ajout d'une colonne :montant_indemnite_modif pour ne pas modifier les valeurs de la colonne MONTANT_INDEMNITÉ_VERSÉ\n\n\nmean_industry = lesion.groupby('INDUSTRIE')['MONTANT_INDEMNITÉ_VERSÉ'].mean() #regroupage des donnés de lesion par INDUSTRIE + calcul la moyenne de ['MONTANT_INDEMNITÉ_VERSÉ'] par industrie\n#print(mean_industry)\nmean_dic = mean_industry.to_dict() #transformer le groupby en dictionnaire\n#print(mean_dic)\n\nfor index, row in lesion.iterrows(): #parcour de lesion\n if pd.isna(row['montant_indemnite_modif']): #check si c'est une valeur manquante\n industrie = row['INDUSTRIE'] #donne a industrie la valeur de la colonne \"INDUSTRIE\n lesion.at[index, 'montant_indemnite_modif'] = mean_dic[industrie] #ajoute la valeur manquante de la colonne \"montant_indemnite_modif\"\n\n# calcul de la moyenne de la nouvelle variable montant_indemnite_modif + arrondissement à la 2eme déciamle\nmontant_indemnite_modif = round(lesion['montant_indemnite_modif'].mean(), 2)\n\n\nprint(montant_indemnite_modif)\n# -\n\n# Question 8 : Utilisez le fichier lesions.xlsx et cause.csv pour répondre à la question.\n#\n# Combien y a-t-il de valeur extrême dans la variable AGE? \n#\n# Ces valeurs sont-elles aberrantes ? Justifiez votre réponse.\n\n# +\nlesion = pd.read_excel(\"lesions.xlsx\")\n\n# calcul des différents percentiles\npremier_quartile = np.percentile(lesion['AGE'], 25)\ntroisieme_quartile = np.percentile(lesion['AGE'], 75.0)\n\n# calcul de l'écart interquartile\necart_interquartile = troisieme_quartile - premier_quartile\n\n# calcul du minimum et du maximum\n\nval_extrem_basse = premier_quartile - 1.5 * ecart_interquartile\nval_extrem_haute = troisieme_quartile + 1.5 * ecart_interquartile\n\nextrem_values = lesion[(lesion['AGE'] < val_extrem_basse) | (lesion['AGE'] > val_extrem_haute)] # | (lesion['AGE'] == 0)]\n# extremvalues = les valeurs inférieur à la valeur minimale attendue ou les valeurs supérieur à la valeur maximale attendue \n#print(extrem_values[\"AGE\"])\nsum_extrem = len(extrem_values)\nprint(sum_extrem)\n# -\n\n# Question 9 :\n#\n# Utilisez le fichier lesions.xlsx et cause.csv pour répondre à la question.\n#\n# Quelle est la moyenne d'âge (arrondie à une décimale) des hommes qui ont eu un accident au mois de Juin?\n\n# +\n\nlesion = pd.read_excel(\"lesions.xlsx\")\n\n#converti DATE_ACCIDENT en datetime\nlesion['DATE_ACCIDENT'] = pd.to_datetime(lesion['DATE_ACCIDENT'], format='%d/%m/%Y')\n\n# recuperer les accidents de Juin\njune_accident = lesion[lesion['DATE_ACCIDENT'].dt.month == 6]\n\n#recuperer les accidents impliquant des hommes\naccident_men_june = june_accident[june_accident['DESCR_SEX_PERS_PHYS'] == 'HOMME']\n\n# Calcul de la moyenne d'âge des hommes en juin + arrondissement à 1 décimale\nmoyenne_age_hommes_juin = round(accident_men_june['AGE'].mean(), 1)\n\nprint(moyenne_age_hommes_juin)\n# -\n\n# question 10 :\n# Créez un DataFrame nommé jour_arret qui présente la moyenne du nombre de jour d'arrêt par genre et selon les 3 groupes d'âge suivants:\n# 25 ans et moins,\n# 25 à 50 ans\n# 50 ans +. \n# Les valeurs de ce Dataframe doivent être arrondie à la deuxième décimale.\n# Copiez le code qui produit ce Dataframe dans le champs pour la réponse ci-dessous.\n\n# +\njour_arret = pd.read_excel(\"lesions.xlsx\")\n\nbins_age = [0, 25, 50, float(\"inf\")] #initialisation des intervals\nbins_labels = [\"25 ans et moins\", \"25 à 50 ans\", \"50 ans +\"] #iniatilisation des étiquettes\n\njour_arret['GROUPE_AGE'] = pd.cut(jour_arret['AGE'], bins=bins_age, labels=bins_labels) #creation d'une nouvelle colonne GROUPE_AGE\n\n\njour_arret = round(jour_arret.groupby(['DESCR_SEX_PERS_PHYS', 'GROUPE_AGE'])['JOUR_ARRET_TRAVAIL'].mean().reset_index(), 2) \n#regroupement des données par sexe, catégorie d'âge, + calcul de la moyenne des jours d'arrêt de travail pour chaque groupe puis arrondissement à la deuxieme décimale\n\nprint(jour_arret)\n# -\n\n# Question 11 :\n# Utilisez le fichier lesions.xlsx et cause.csv pour répondre à la question.\n#\n# Combien d'accident de travail ont une description de cause qui contient 3 mots ?\n\n# +\ndf = pd.read_csv(\"./cause.csv\")\n\ncount = 0\n#parcour dataframe\nfor index, row in df.iterrows():\n words = row['DESCR_AGEN_CAUS_LESI'].split('|') #division des str par |\n if len(words) == 3: \n count += 1\n\nprint(count)\n# -\n\n# Question 12: Utilisez le fichier lesions.xlsx et cause.csv pour répondre à la question.\n#\n# Combien de jours ce sont écoulés entre l'accident ACC_1062 et l'accident ACC_833?\n\n# +\nlesion = pd.read_excel(\"lesions.xlsx\")\n\n# Convertir la colonne de dates en format de date\nlesion['DATE_ACCIDENT'] = pd.to_datetime(lesion['DATE_ACCIDENT'], format='%d/%m/%Y')\n\n# Recherche de la date de l'accident ACC_1062 avec iloc\ndate_acc_1062 = lesion[lesion['ID'] == 'ACC_1062']['DATE_ACCIDENT'].iloc[0]\n#print(date_acc_1062)\n\ndate_acc_833 = lesion[lesion['ID'] == 'ACC_833']['DATE_ACCIDENT'].iloc[0]\n#print(date_acc_833)\n\n# Calcul de la différence entre les deux dates\ndays = (date_acc_1062 - date_acc_833).days\n\nprint(days)\n# -\n\n# Question 13: Utilisez le fichier lesions.xlsx et cause.csv pour répondre à la question.\n#\n# Quelle est l'industrie avec le plus grand ratio d'accidents impliquant des hommes et des femmes?\n#\n# Le ratio d'accidents se calcule en divisant le nombre d'accidents impliquant des hommes par le nombre d'accident impliquant des femmes. \n#\n# Ne considérez pas les industries qui n'ont que des femmes ou que des hommes impliqués dans leur accident de travail. \n\n# +\nimport pandas as pd\n\nlesions = pd.read_excel('lesions.xlsx')\n\n# Filtre les industries avec des accidents d'hommes et de femmes\ndef filter_industries(group):\n contains_homme = \"HOMME\" in group['DESCR_SEX_PERS_PHYS'].values\n contains_femme = \"FEMME\" in group['DESCR_SEX_PERS_PHYS'].values\n return contains_homme and contains_femme\n\nfiltered = lesions.groupby('INDUSTRIE').filter(filter_industries)\n\n# Creation d'une table pivot pour le nombre total d'accidents par genre et par industrie\npivot_table = pd.pivot_table(filtered, values='ID', index=['INDUSTRIE'], columns=['DESCR_SEX_PERS_PHYS'], aggfunc='count', fill_value=0)\n#Calcul du ratio\npivot_table['Ratio_Homme_Femme'] = pivot_table['HOMME'] / pivot_table['FEMME']\n#avoir le ratio maximum grace a idxmax\nindice_max_ratio = pivot_table['Ratio_Homme_Femme'].idxmax()\n\nprint(indice_max_ratio)\n\n\n# -\n\n# Question 14 : Utilisez le fichier lesions.xlsx et cause.csv pour répondre à la question.\n#\n# Quelle est la moyenne (arrondie à deux décimales) d'âge des victimes d'accidents dont la valeur est manquante pour la variable MONTANT_INDEMNITÉ_VERSÉ ?\n\n# +\nlesion = pd.read_excel(\"lesions.xlsx\")\n\n#mis dans missing les lignes du DataFrame lesion où la colonne 'MONTANT_INDEMNITÉ_VERSÉ' est nulle\nmissing = lesion[lesion['MONTANT_INDEMNITÉ_VERSÉ'].isnull()]\n\nmissing_mean = round(missing['AGE'].mean(), 2)\n\nprint(missing_mean)\n","repo_name":"matt-tek/data-science-1","sub_path":"WalgraefGabrielleWALG13580100.ipynb","file_name":"WalgraefGabrielleWALG13580100.ipynb","file_ext":"py","file_size_in_byte":11798,"program_lang":"python","lang":"fr","doc_type":"code","stars":0,"dataset":"github-jupyter-script","pt":"44"}
+{"seq_id":"72689816772","text":"import pandas as pd\nimport numpy as np\nimport matplotlib.pyplot as ply\n\n#load data\ncell_df = pd.read_csv('cell_samples.csv')\ncell_df.head()\n\n# +\n#distribution of classes\nbenign_df = cell_df[cell_df['Class']==2][0:200]\nmalignant_df = cell_df[cell_df['Class']==4][0:200]\n#help(benign_df.plot)\naxes = benign_df.plot(kind = 'scatter',x = 'Clump',y = 'UnifSize',color = 'blue',label = 'Benign')\n\nmalignant_df.plot(kind = 'scatter',x = 'Clump',y = 'UnifSize',color = 'red',label = 'malignant',ax=axes)\n# -\n\n#unwanted rows\ncell_df.dtypes\n\ncell_df = cell_df[pd.to_numeric(cell_df['BareNuc'],errors = 'coerce').notnull()]\ncell_df['BareNuc'] = cell_df['BareNuc'].astype('int')\ncell_df.dtypes\n\n# +\n#remove unwanted columns\ncell_df.columns\nfeature_df = cell_df[['Clump', 'UnifSize', 'UnifShape', 'MargAdh', 'SingEpiSize',\n 'BareNuc', 'BlandChrom', 'NormNucl', 'Mit']]\n\n#numpy nd array, independent array\nX = np.asarray(feature_df)\n#dependent array\ny = np.asarray(cell_df['Class'])\nX[0:5]\n# -\n\n#train and test set\nfrom sklearn.model_selection import train_test_split\nX_train,X_test,y_train,y_test = train_test_split(X,y,test_size = 0.2,random_state = 4)\nX_train.shape\ny_train.shape\n\nfrom sklearn import svm\nclassifier = svm.SVC(kernel='linear', gamma = 'auto', C=2)\nclassifier.fit(X_train,y_train)\ny_predict = classifier.predict(X_test)#1D\n\n\n#result\nfrom sklearn.metrics import classification_report\nprint(classification_report(y_test,y_predict))\n","repo_name":"AfTeReFfEcT1/SVMmodels_imageProcessing","sub_path":"SVMmodels_image_processing/SVM_yt_breast_cancer.ipynb","file_name":"SVM_yt_breast_cancer.ipynb","file_ext":"py","file_size_in_byte":1440,"program_lang":"python","lang":"en","doc_type":"code","stars":0,"dataset":"github-jupyter-script","pt":"44"}
+{"seq_id":"33216400798","text":"# # ELEC474 Prelab4\n#\n#\n\nimport cv2 as cv\nimport numpy as np\nimport random\nimport math\nimport time\nfrom matplotlib import pyplot as plt\n\n# +\nglobal my_SIFT_instance, my_BF_instance\nmy_SIFT_instance = cv.SIFT_create()\nmy_BF_instance = cv.BFMatcher()\n\n\nimg1Name = \"lena.png\"\nimg2Name = \"backpack_left.png\"\nimg3Name = \"backpack_right.png\"\nimg4Name = \"selfTest1.jpg\"\nimg5Name = \"selfTest2.jpg\"\n\nimgDescipt_1 = np.array((\n\"lena.png keypoint\",\n\"backpack_left.png keypoint\",\n\"backpack_right.png keypoint\"\n))\n\nimgDescipt_2 = np.array((\n \"Lena Match\",\n \"Backpack Match\"\n))\n\n# -\n\n# ## 1.1\n\ndef PltImg(img,imgDescipt):\n plt.figure(dpi=300)\n plt.figure(figsize=(15,15))\n idx = len(img)\n for i in range(len(img)):\n plt.subplot(1,idx,i+1)\n plt.imshow(img[i],cmap=\"gray\")\n plt.title(imgDescipt[i])\n plt.tight_layout()\n\n\n# +\n\ndef SIFTOut(imgName):\n img = cv.imread(imgName)\n imgGray = cv.cvtColor(img,cv.COLOR_BGR2GRAY)\n imgKpOut = np.copy(imgGray)\n kp, des = my_SIFT_instance.detectAndCompute(imgGray,None)\n imgKpOut = cv.drawKeypoints(imgKpOut,kp,0)\n return imgKpOut,kp,des\n\n\n# -\n\ndef BFMatchDescriptor(pram1, pram2, kNum):\n des1 = pram1[2]\n des2 = pram2[2]\n matches = my_BF_instance.knnMatch(des1,des2,k = kNum)\n return cv.drawMatchesKnn(pram1[0],pram1[1],pram2[0],pram2[1],matches,None,flags=2)\n\n\n# +\noutImages = []\nmatchImages = []\noutImages.append(SIFTOut(img1Name))\noutImages.append(SIFTOut(img2Name))\noutImages.append(SIFTOut(img3Name))\noutImages = np.array(outImages,dtype=object)\n\nmatchImages.append(BFMatchDescriptor(outImages[0],outImages[0],2))\nmatchImages.append(BFMatchDescriptor(outImages[1],outImages[2],2))\n\nPltImg(outImages[:,0],imgDescipt_1)\nPltImg(matchImages,imgDescipt_2)\n\n\n# -\n\n# ## 1.2\n\ndef BFMatchLoweRatio(pram1, pram2, kNum, ratio):\n matches = my_BF_instance.knnMatch(pram1[2],pram2[2],k = kNum)\n good = []\n goodDis = []\n allMatchDis = []\n for m,n in matches:\n diff = abs(m.distance - n.distance)\n if m.distance < ratio * n.distance:\n good.append([m])\n goodDis.append(m.distance)\n allMatchDis.append(m.distance)\n return [cv.drawMatchesKnn(pram1[0],pram1[1],pram2[0],pram2[1],good,None,flags=2),allMatchDis,goodDis]\n\n\n# +\nmatchRatioImage = []\nmatchRatioImage.append(BFMatchLoweRatio(outImages[0],outImages[0],2,0.8))\nmatchRatioImage.append(BFMatchLoweRatio(outImages[1],outImages[2],2,0.8))\nmatchRatioImage = np.array(matchRatioImage,dtype=object)\n\nPltImg(matchRatioImage[:,0],imgDescipt_2)\n\n\n# -\n\n# ## 1.3\n\ndef HistogramCompare(allMatch, loweMatch, GroupNum):\n plt.figure(dpi=75)\n\n plt.hist(allMatch, bins = GroupNum, label=\"All Matchs\")\n plt.hist(loweMatch,bins = GroupNum, label=\"Lowe's Matches\")\n plt.xlabel(\"Distance between the two descriptors\")\n plt.ylabel(\"Count of the histograms with that distance\")\n plt.legend()\n plt.show()\n \n\n\nHistogramCompare(matchRatioImage[0,1],matchRatioImage[0,2],300)\nHistogramCompare(matchRatioImage[1,1],matchRatioImage[1,2],300)\n\n# ## Q1 Question: \n# ### What conclusion can be drawn from this histogram?\n\n# ## Answer\n# ### The lowe's Matches sucessfuly filter out the points that are \"good\". Since descriptor is in relatively high dimensional space, the distance will reflect how close those two points are, the smaller the value, the better the result. and since lena has no move at all (which use the same image for compare), the distance are 0\n# ### Furthermore, the image below using more restrict filter to the matches, which result in significant drop on matches number\n\n# +\nmatchRatioImage = []\nmatchRatioImage.append(BFMatchLoweRatio(outImages[0],outImages[0],2,0.4))\nmatchRatioImage.append(BFMatchLoweRatio(outImages[1],outImages[2],2,0.4))\nmatchRatioImage = np.array(matchRatioImage,dtype=object)\n\nPltImg(matchRatioImage[:,0],imgDescipt_2)\nHistogramCompare(matchRatioImage[0,1],matchRatioImage[0,2],300)\nHistogramCompare(matchRatioImage[1,1],matchRatioImage[1,2],300)\n","repo_name":"OuNiSang/ELEC474_Lab_Backup","sub_path":"Prelab4/Main_MikeFeng_20119641.ipynb","file_name":"Main_MikeFeng_20119641.ipynb","file_ext":"py","file_size_in_byte":3986,"program_lang":"python","lang":"en","doc_type":"code","stars":0,"dataset":"github-jupyter-script","pt":"44"}
+{"seq_id":"38363979335","text":"# # Исследование объявлений о продаже квартир\n#\n# В вашем распоряжении данные сервиса Яндекс.Недвижимость — архив объявлений о продаже квартир в Санкт-Петербурге и соседних населённых пунктов за несколько лет. Нужно научиться определять рыночную стоимость объектов недвижимости. Ваша задача — установить параметры. Это позволит построить автоматизированную систему: она отследит аномалии и мошенническую деятельность. \n#\n# По каждой квартире на продажу доступны два вида данных. Первые вписаны пользователем, вторые — получены автоматически на основе картографических данных. Например, расстояние до центра, аэропорта, ближайшего парка и водоёма. \n\n# ### Откройте файл с данными и изучите общую информацию. \n\n# +\nimport pandas as pd\nimport numpy as np\nimport matplotlib.pyplot as plt\nimport seaborn\nfrom IPython.display import display\n\n# Запишем датафрейм в переменную и выведем первые 15 строк для ознакомления. С той же целью вызовем функцию info()\n\ndf = pd.read_csv(\"/datasets/real_estate_data.csv\", sep = '\\t')\ndisplay(df.head(15))\ndf.info()\n# -\n\n# In this exercise, you will analyze a dataset of human body temperatures and employ the concepts of hypothesis testing, confidence intervals, and statistical significance.
\n#\n#Answer the following questions in this notebook below and submit to your Github account.
\n#\n#\n\n# + id=\"TXhTgIYvL2BF\"\nimport requests, codecs\nfrom bs4 import BeautifulSoup\nimport pandas as pd\n\n# + id=\"ca0262U7R0le\"\ndf = pd.DataFrame(columns=[\"地區\",\"氣溫\"])\n\n# + id=\"71ZMiq_0Lk8a\"\nurl = requests.get(\"https://www.cwb.gov.tw/V8/C/W/TemperatureTop/County_TMax_T.html\")\nreq = BeautifulSoup(url.text,\"html.parser\")\ndata = req.find_all(\"tr\")\nfor d in data:\n city = d.find(\"th\").text\n temp = d.find(\"span\",{\"class\" : \"tem-C\"}).text\n s = pd.Series([city, temp],index=[\"地區\",\"氣溫\"])\n df = df.append(s,ignore_index=True)\n\n# + id=\"ceN-udgKTgLl\" outputId=\"9527342d-70ce-41e2-bc2f-a78880daa6e7\" colab={\"base_uri\": \"https://localhost:8080/\", \"height\": 739}\ndf\n","repo_name":"syuan0525/crawler","sub_path":"crawler_HW3.ipynb","file_name":"crawler_HW3.ipynb","file_ext":"py","file_size_in_byte":929,"program_lang":"python","lang":"en","doc_type":"code","stars":0,"dataset":"github-jupyter-script","pt":"44"}
+{"seq_id":"73060284934","text":"# This code is used to analyze the correlation between different parameters\n\n# +\nimport seaborn as sns\nimport numpy as np\nimport matplotlib.pyplot as plt\nimport pandas as pd\n\ndata = pd.read_csv('density_data.csv') \nx = data[['energy_density', 'size']]\ny = data['porosity']\nz = data[['energy_density', 'size','speed','density']]\nz.columns=['Energy density', 'Geometry', 'Laser scanning speed', 'Density']\nsns.pairplot(z)\n\n\n# +\n\nfigure, ax = plt.subplots(figsize=(12, 12))\nsns.heatmap(z.corr(), square=True, annot=True, ax=ax)\n# -\n\n# Calculating MIC\n\n# +\nfrom minepy import MINE\n\ndef MIC_matirx(dataframe, mine):\n\n data = np.array(dataframe)\n n = len(data[0, :])\n result = np.zeros([n, n])\n\n for i in range(n):\n for j in range(n):\n mine.compute_score(data[:, i], data[:, j])\n result[i, j] = mine.mic()\n result[j, i] = mine.mic()\n RT = pd.DataFrame(result)\n return RT\n\nmine = MINE(alpha=0.6, c=15)\ndata_wine_mic = MIC_matirx(z, mine)\ndata_wine_mic\n\n","repo_name":"lzqiii/Prediction-of-Part-Properties-in-PBF-LB-P-of-Polymers-Using-IR-Thermography","sub_path":"4 Correlation Analysis/colleration analysis.ipynb","file_name":"colleration analysis.ipynb","file_ext":"py","file_size_in_byte":1007,"program_lang":"python","lang":"en","doc_type":"code","stars":1,"dataset":"github-jupyter-script","pt":"44"}
+{"seq_id":"8653346457","text":"# ### 1. Найти интервалы возрастания и убывания функций:\n#\n# \n\n# + [markdown] id=\"yvKxvKLfv9Ys\"\n# ___\n#\n# # WELCOME!\n#\n# ___\n\n# + [markdown] id=\"d-_l2fuwv9Yw\"\n# Welcome to the \"***Sentiment Analysis and Classification***\" study.\n#\n# This analysis will focus on using Natural Language techniques to find broad trends in the written thoughts of the customers. \n# The goal is to predict whether customers recommend the product they purchased using the information in their review text.\n#\n# One of the challenges in this study is to extract useful information from the *Review Text* variable using text mining techniques. The other challenge is that we need to convert text files into numeric feature vectors to run machine learning algorithms.\n#\n# We will build sentiment classification models using Machine Learning algorithms (***Logistic Regression, Naive Bayes, Support Vector Machine, Random Forest*** and ***Ada Boosting***) and **Deep Learning algorithms**.\n#\n# Before diving into the project, let's take a look at the Determines and Tasks.\n\n# + [markdown] id=\"N-54I9m9v9Yy\"\n# ## Determines\n#\n# The data is a collection of 22641 Rows and 10 column variables. Each row includes a written comment as well as additional customer information. \n# Also each row corresponds to a customer review, and includes the variables:\n#\n#\n# **Feature Information:**\n#\n# **Clothing ID:** Integer Categorical variable that refers to the specific piece being reviewed.\n#\n# **Age:** Positive Integer variable of the reviewers age.\n#\n# **Title:** String variable for the title of the review.\n#\n# **Review Text:** String variable for the review body.\n#\n# **Rating:** Positive Ordinal Integer variable for the product score granted by the customer from 1 Worst, to 5 Best.\n#\n# **Recommended IND:** Binary variable stating where the customer recommends the product where 1 is recommended, 0 is not recommended.\n#\n# **Positive Feedback Count:** Positive Integer documenting the number of other customers who found this review positive.\n#\n# **Division Name:** Categorical name of the product high level division.\n#\n# **Department Name:** Categorical name of the product department name.\n#\n# **Class Name:** Categorical name of the product class name.\n#\n# ---\n#\n# The basic goal in this project is to predict whether customers recommend the product they purchased using the information in their *Review Text*.\n# Especially, it should be noted that the expectation in this project is to use only the \"Review Text\" variable and neglect the other ones. \n# Of course, if we want, we can work on other variables individually.\n#\n# Project Structure is separated in five tasks: ***EDA, Feature Selection and Data Cleaning , Text Mining, Word Cloud*** and ***Sentiment Classification with Machine Learning***.\n#\n# Classically, we can start to know the data after doing the import and load operations. \n# We need to do missing value detection for Review Text, which is the only variable we need to care about. We can drop other variables.\n#\n# We will need to apply ***noise removal*** and ***lexicon normalization*** processes by using the capabilities of the ***nltk*** library to the data set that is ready for text mining.\n#\n# Afterwards, we will implement ***Word Cloud*** as a visual analysis of word repetition.\n#\n# Finally, we will build models with five different algorithms and compare their performance. Thus, you will determine the algorithm that makes the most accurate emotion estimation by using the information obtained from the * Review Text * variable.\n#\n#\n#\n#\n#\n\n# + [markdown] id=\"igR2ru-Ev9Yz\"\n# ---\n# ---\n\n# + [markdown] id=\"CKEu_C_8v9Yz\"\n# ## Tasks\n#\n# #### 1. Exploratory Data Analysis\n#\n# - Import Modules, Load Discover the Data\n#\n# #### 2. Feature Selection and Data Cleaning\n#\n# - Feature Selection and Rename Column Name\n# - Missing Value Detection\n#\n# #### 3. Text Mining\n#\n# - Tokenization\n# - Noise Removal\n# - Lexicon Normalization\n#\n# #### 4. WordCloud - Repetition of Words\n#\n# - Detect Reviews\n# - Collect Words \n# - Create Word Cloud \n#\n#\n# #### 5. Sentiment Classification with Machine Learning\n#\n# - Train - Test Split\n# - Vectorization\n# - TF-IDF\n# - Logistic Regression\n# - Naive Bayes\n# - Support Vector Machine\n# - Random Forest\n# - AdaBoost\n# - Model Comparison\n\n# + [markdown] id=\"4STsyYu7v9Y0\"\n# ---\n# ---\n\n# + [markdown] id=\"HOkP9vyYCV31\"\n# # Sentiment analysis of women's clothes reviews\n#\n#\n# In this study we used sentiment analysis to determined whether the product is recommended or not. We used different machine learning algorithms to get more accurate predictions. The following classification algorithms have been used: Logistic Regression, Naive Bayes, Support Vector Machine (SVM), Random Forest and Ada Boosting. The dataset comes from Woman Clothing Review that can be find at (https://www.kaggle.com/nicapotato/womens-ecommerce-clothing-reviews. \n#\n\n# + id=\"a9bgl5gZv9Y1\"\n# This Python 3 environment comes with many helpful analytics libraries installed\n# It is defined by the kaggle/python Docker image: https://github.com/kaggle/docker-python\n# For example, here's several helpful packages to load\n\nimport numpy as np # linear algebra\nimport pandas as pd # data processing, CSV file I/O (e.g. pd.read_csv)\n\n# Input data files are available in the read-only \"../input/\" directory\n# For example, running this (by clicking run or pressing Shift+Enter) will list all files under the input directory\n\nimport os\nfor dirname, _, filenames in os.walk('/kaggle/input'):\n for filename in filenames:\n print(os.path.join(dirname, filename))\n\n# You can write up to 20GB to the current directory (/kaggle/working/) that gets preserved as output when you create a version using \"Save & Run All\" \n# You can also write temporary files to /kaggle/temp/, but they won't be saved outside of the current session\n\n# + [markdown] id=\"HtUukhVRv9Y2\"\n# ## 1. Exploratory Data Analysis\n\n# + [markdown] id=\"FXzFEeYkCV32\"\n# ### Import Libraries, Load and Discover the Data\n\n# + colab={\"base_uri\": \"https://localhost:8080/\", \"height\": 104} id=\"65J2M9pav9Y2\" outputId=\"196d4317-5103-47b5-b2c6-98a6772f27a9\"\n# # !pip install pyforest\n# 1-Import Libraies\nimport numpy as np\nimport pandas as pd \nimport seaborn as sns\nimport matplotlib.pyplot as plt\nimport scipy.stats as stats\n# %matplotlib inline\nimport statsmodels.api as sm\nimport statsmodels.formula.api as smf\nimport missingno as msno \n\nfrom sklearn.compose import make_column_transformer\n\n# Scaling\nfrom sklearn.preprocessing import scale \nfrom sklearn.preprocessing import StandardScaler\nfrom sklearn.preprocessing import PolynomialFeatures \nfrom sklearn.preprocessing import OneHotEncoder\nfrom sklearn.preprocessing import PowerTransformer \nfrom sklearn.preprocessing import MinMaxScaler\nfrom sklearn.preprocessing import RobustScaler\n\n\n# Importing plotly and cufflinks in offline mode\nimport plotly.express as px\nimport cufflinks as cf\nimport plotly.offline\ncf.go_offline()\ncf.set_config_file(offline=False, world_readable=True)\n\n# Ignore Warnings\nimport warnings\nwarnings.filterwarnings(\"ignore\")\nwarnings.warn(\"this will not show\")\n\n# Figure&Display options\nplt.rcParams[\"figure.figsize\"] = (10,6)\npd.set_option('max_colwidth',200)\npd.set_option('display.max_rows', 1000)\npd.set_option('display.max_columns', 200)\npd.set_option('display.float_format', lambda x: '%.3f' % x)\n\n# # !pip install termcolor\n# !pip install colorama\nimport colorama\nfrom colorama import Fore, Style # maakes strings colored\nfrom termcolor import colored\n\nimport ipywidgets\nfrom ipywidgets import interact\n\nimport nltk\nnltk.download('punkt')\nnltk.download(\"stopwords\")\nnltk.download('all')\nfrom nltk.tokenize import sent_tokenize, word_tokenize\nfrom nltk.corpus import stopwords\nfrom nltk.stem import WordNetLemmatizer\nfrom nltk.stem import PorterStemmer\nfrom collections import Counter\nfrom wordcloud import WordCloud \n\n\n# + [markdown] id=\"9U7Hp8frv9Y3\"\n# ### Some Useful User Defined Functions\n\n# + id=\"3O2TpkXtv9Y3\"\n## Some Useful Functions\n\n###############################################################################\n\ndef missing_values(df):\n missing_number = df.isnull().sum().sort_values(ascending=False)\n missing_percent = (df.isnull().sum()/df.isnull().count()).sort_values(ascending=False)\n missing_values = pd.concat([missing_number, missing_percent], axis=1, keys=['Missing_Number', 'Missing_Percent'])\n return missing_values[missing_values['Missing_Number']>0]\n\n###############################################################################\n\ndef first_looking(df):\n print(colored(\"Shape:\", attrs=['bold']), df.shape,'\\n', \n colored('-'*79, 'red', attrs=['bold']),\n colored(\"\\nInfo:\\n\", attrs=['bold']), sep='')\n print(df.info(), '\\n', \n colored('-'*79, 'red', attrs=['bold']), sep='')\n print(colored(\"Number of Uniques:\\n\", attrs=['bold']), df.nunique(),'\\n',\n colored('-'*79, 'red', attrs=['bold']), sep='')\n print(colored(\"Missing Values:\\n\", attrs=['bold']), missing_values(df),'\\n', \n colored('-'*79, 'red', attrs=['bold']), sep='')\n print(colored(\"All Columns:\", attrs=['bold']), list(df.columns),'\\n', \n colored('-'*79, 'red', attrs=['bold']), sep='')\n\n df.columns= df.columns.str.lower().str.replace('&', '_').str.replace(' ', '_')\n\n print(colored(\"Columns after rename:\", attrs=['bold']), list(df.columns),'\\n',\n colored('-'*79, 'red', attrs=['bold']), sep='')\n \n \ndef multicolinearity_control(df):\n feature =[]\n collinear=[]\n for col in df.corr().columns:\n for i in df.corr().index:\n if (abs(df.corr()[col][i])> .9 and abs(df.corr()[col][i]) < 1):\n feature.append(col)\n collinear.append(i)\n print(colored(f\"Multicolinearity alert in between:{col} - {i}\", \n \"red\", attrs=['bold']), df.shape,'\\n',\n colored('-'*79, 'red', attrs=['bold']), sep='')\n\ndef duplicate_values(df):\n print(colored(\"Duplicate check...\", attrs=['bold']), sep='')\n duplicate_values = df.duplicated(subset=None, keep='first').sum()\n if duplicate_values > 0:\n df.drop_duplicates(keep='first', inplace=True)\n print(duplicate_values, colored(\"Duplicates were dropped!\"),'\\n',\n colored('-'*79, 'red', attrs=['bold']), sep='')\n else:\n print(colored(\"There are no duplicates\"),'\\n',\n colored('-'*79, 'red', attrs=['bold']), sep='') \n \ndef drop_columns(df, drop_columns):\n if drop_columns !=[]:\n df.drop(drop_columns, axis=1, inplace=True)\n print(drop_columns, 'were dropped')\n else:\n print(colored('We will now check the missing values and if necessary will drop related columns!', attrs=['bold']),'\\n',\n colored('-'*79, 'red', attrs=['bold']), sep='')\n \ndef drop_null(df, limit):\n print('Shape:', df.shape)\n for i in df.isnull().sum().index:\n if (df.isnull().sum()[i]/df.shape[0]*100)>limit:\n print(df.isnull().sum()[i], 'percent of', i ,'null and were dropped')\n df.drop(i, axis=1, inplace=True)\n print('new shape:', df.shape) \n print('New shape after missing value control:', df.shape)\n \n###############################################################################\n\n# To view summary information about the column\n\ndef first_look(col):\n print(\"column name : \", col)\n print(\"--------------------------------\")\n print(\"per_of_nulls : \", \"%\", round(df[col].isnull().sum()/df.shape[0]*100, 2))\n print(\"num_of_nulls : \", df[col].isnull().sum())\n print(\"num_of_uniques : \", df[col].nunique())\n print(df[col].value_counts(dropna = False))\n\n\n# + id=\"G_4Unah5v9Y4\"\ndf = pd.read_csv(\"Womens Clothing E-Commerce Reviews (1).csv\")\n\n# + id=\"V-hE-_jBv9Y4\"\n#df = pd.read_csv(\"../input/womens-ecommerce-clothing-reviews/Womens Clothing E-Commerce Reviews.csv\")\n\n# + colab={\"base_uri\": \"https://localhost:8080/\", \"height\": 649} id=\"J9iKVU8vv9Y5\" outputId=\"00921e15-2a27-4ebc-9183-36ef615db101\"\ndf.tail()\n\n# + colab={\"base_uri\": \"https://localhost:8080/\", \"height\": 684} id=\"6boAgnXmv9Y5\" outputId=\"72ad3565-bed6-408c-c3df-956da7b32bed\"\ndf.sample(5)\n\n# + [markdown] id=\"T54GVJwICV36\"\n# ### Data Wrangling\n\n# + colab={\"base_uri\": \"https://localhost:8080/\"} id=\"EMr5-6Adv9Y5\" outputId=\"955dfe72-e446-4c7d-cc9a-fa24d4b6fbd2\"\nfirst_looking(df)\nduplicate_values(df)\ndrop_columns(df,[])\ndrop_null(df, 90)\n\n# + colab={\"base_uri\": \"https://localhost:8080/\"} id=\"z6aKVu8xv9Y6\" outputId=\"5a1f1a1b-7d32-4449-c946-4f62fc8b2423\"\ndf.columns\n\n# + [markdown] id=\"dg3em4YFv9Y6\"\n# - The \"Unnamed:_0\" column contains completely unique values and contains the same information as the index. Alsa \"clothing_id\" columns has unique values over 1200. I'm dropping this columns because they won't work for us as they stand.\n\n# + id=\"R8o-ixKQv9Y6\"\ndf.drop(['unnamed:_0', 'clothing_id'], axis = 1, inplace=True)\n\n# + colab={\"base_uri\": \"https://localhost:8080/\"} id=\"OFLOQ_Nxv9Y7\" outputId=\"eef25396-8353-4a1b-fc30-9363418306bb\"\ndf.shape\n\n# + id=\"w5b6ojWTv9Y7\"\ndf = df.rename(columns = {'Review Text' : 'text', 'recommended_ind' : 'recommended', \n 'positive_feedback_count' : 'feedback_count', 'division_name' : 'division', \n 'department_name' : 'department', 'class_name' :'class'})\n\n# + colab={\"base_uri\": \"https://localhost:8080/\", \"height\": 174} id=\"HD7TxawVv9Y7\" outputId=\"7725bab4-89c2-437f-a64c-c2269aa5798d\"\ndf.describe().T\n\n# + [markdown] id=\"x2qja5k5v9Y7\"\n# - Columns with ordinal information, although the \"rating\" and \"recommended_ind\" columns are encoded as numeric. \n\n# + colab={\"base_uri\": \"https://localhost:8080/\", \"height\": 206} id=\"apFqT_vov9Y8\" outputId=\"0ab547e9-3a9d-4de7-c345-d698335b5731\"\ndf.describe(include=object).T\n\n# + colab={\"base_uri\": \"https://localhost:8080/\", \"height\": 379} id=\"BYn75zTjv9Y8\" outputId=\"abc930bd-b46c-4658-8842-01aedcb41e07\"\nsns.heatmap(df.corr(), annot=True);\n\n# + [markdown] id=\"Fk1f6AIGv9Y8\"\n# - Let us have look at the columns remaining in the dataset.\n\n# + colab={\"base_uri\": \"https://localhost:8080/\"} id=\"u_78-6dDv9Y9\" outputId=\"551081dd-dbec-4dd2-d6cd-c25f30f904db\"\ndf.columns\n\n# + [markdown] id=\"5VvNZS1Iv9Y9\"\n# ***age***\n# - Positive Integer variable of the reviewers age.\n\n# + colab={\"base_uri\": \"https://localhost:8080/\"} id=\"-A-PytpUv9Y-\" outputId=\"a76c9014-1c24-4ac7-e883-0294b491f375\"\nfirst_look(\"age\")\n\n# + id=\"gQ7VsSNiv9Y-\"\ndf.age.describe().T\n\n# + id=\"BtmrMDCqv9Y-\"\npx.histogram(df, x = df.age)\n\n# + id=\"ZqpWgtQNv9Y_\"\npd.crosstab(df.age, df.recommended).iplot(kind=\"bar\")\n\n# + [markdown] id=\"8DNknrzXv9Y_\"\n# ***rating*** \n# - Positive Ordinal Integer variable for the product score granted by the customer from 1 Worst, to 5 Best.\n\n# + id=\"WAh0zJAgv9Y_\"\nfirst_look(\"rating\")\n\n# + id=\"KNwSPwBlv9ZA\"\ndf.rating.describe().T\n\n# + id=\"EoQlD6wBv9ZA\"\npx.histogram(df, x = df.rating)\n\n# + id=\"WL1VP1RMv9ZA\"\npd.crosstab(df.rating, df.recommended).iplot(kind=\"bar\")\n\n# + [markdown] id=\"HRuXIg2Yv9ZA\"\n# ***recommended*** \n# - Binary variable stating where the customer recommends the product where 1 is recommended, 0 is not recommended.\n\n# + id=\"efpeFXR7v9ZB\"\nfirst_look(\"recommended\")\n\n# + id=\"GHOP7NCKv9ZB\"\ndf.recommended.describe().T\n\n# + id=\"uWrjLJyHv9ZB\"\npx.histogram(df, x = df.recommended)\n\n# + [markdown] id=\"_6XTM13pv9ZB\"\n# ***feedback_count:***\n# - Positive Integer documenting the number of other customers who found this review positive.\n\n# + id=\"Htf0aA9Bv9ZB\"\nfirst_look(\"feedback_count\")\n\n# + id=\"PPjGsDZiv9ZC\"\ndf.feedback_count.describe().T\n\n# + id=\"xQMAJs7tv9ZC\"\npx.histogram(df, x = df.feedback_count)\n\n# + id=\"W4LZ8P3sv9ZC\"\npd.crosstab(df.feedback_count, df.recommended).iplot(kind=\"bar\")\n\n# + [markdown] id=\"l2MNXcGqv9ZC\"\n# ***division*** \n# - Categorical name of the product high level division.\n\n# + id=\"oU1c8YQvv9ZD\"\nfirst_look(\"division\")\n\n# + id=\"yUZxEP5yv9ZD\"\ndf.division.describe().T\n\n# + id=\"fuWobPqZv9ZD\"\npx.histogram(df, x = df.division)\n\n# + id=\"YpKY8teOv9ZD\"\npd.crosstab(df.division, df.recommended).iplot(kind=\"bar\")\n\n# + [markdown] id=\"YVfzKYNtv9ZD\"\n# ***department***\n# - Categorical name of the product department name\n\n# + id=\"quqWqHMcv9ZE\"\nfirst_look(\"department\")\n\n# + id=\"qNMqjcnuv9ZE\"\ndf.department.describe().T\n\n# + id=\"Mr56JXd_v9ZE\"\npx.histogram(df, x = df.department)\n\n# + id=\"Y2Ctkr3ov9ZE\"\npd.crosstab(df.department, df.recommended).iplot(kind=\"bar\")\n\n# + [markdown] id=\"r0wAlgwgv9ZF\"\n# ***class***\n# - Categorical name of the product class name.\n\n# + id=\"OMC3tGGBv9ZL\"\nfirst_look(\"class\")\n\n# + id=\"qHmQmr9Av9ZN\"\ndf[\"class\"].describe().T\n\n# + id=\"_iOuvAHQv9ZO\"\npx.histogram(df, x = df[\"class\"])\n\n# + id=\"Yz5waRR7v9ZP\"\npd.crosstab(df[\"class\"], df.recommended).iplot(kind=\"bar\")\n\n# + [markdown] id=\"mOe38k7rv9ZQ\"\n# ***title*** \n# - String variable for the title of the review.\n\n# + id=\"B5yoZ6niv9ZQ\"\nfirst_look(\"title\")\n\n# + id=\"x6jpjuZEv9ZQ\"\ndf.title.describe().T\n\n# + [markdown] id=\"GJpgfpx2v9ZQ\"\n# ***review_text*** \n# - String variable for the review body.\n\n# + id=\"eCovngwTv9ZQ\"\nfirst_look(\"review_text\")\n\n# + id=\"NH9w5RgWv9ZR\"\ndf.review_text.describe().T\n\n# + [markdown] id=\"AQuhTpcwCV38\"\n# #### Check Proportion of Target Class Variable:\n\n# + id=\"-FnfY3oev9ZR\"\ndf.columns\n\n# + [markdown] id=\"0YXKZ5QgCV39\"\n# The target class variable is imbalanced, where \"Recommended\" values are more dominating then \"Not Recommended\".\n\n# + id=\"wjHhj9Ayv9ZS\"\n# recommended : \n# Binary variable stating where the customer recommends the product where 1 is recommended, 0 is not recommended.\n\nprint(df.recommended.value_counts())\n\nplt.figure(figsize=(10,10))\n\nexplode = [0,0.1]\nplt.pie(df.recommended.value_counts(), explode=explode,autopct='%1.1f%%', shadow=True,startangle=140)\nplt.legend(labels=['1 : recommended','0 : not recommended'])\nplt.title('Recommendation Distribution')\nplt.axis('off');\n\n# + [markdown] id=\"cPEVV0JGCV37\"\n# ## 2. Feature Selection and Data Cleaning\n#\n# From now on, the DataFrame we will work with should contain two columns: **\"Review Text\"** and **\"Recommended IND\"**. We can do the missing value detection operations from now on. We can also rename the column names if we want.\n\n# + [markdown] id=\"VFu-W4wnv9ZS\"\n# ### Feature Selection and Rename Column Name\n\n# + colab={\"base_uri\": \"https://localhost:8080/\"} id=\"Kv1sPC_Cv9ZS\" outputId=\"00b9b7d4-0409-4e48-facf-83913aa8c2c2\"\ndf.columns\n\n# + id=\"bYM7uPwdv9ZS\"\ndf_ml = df.copy()\n\n# df_ml = df[['review_text', 'recommended']].copy()\n\n# + id=\"zRFqu0dlv9ZT\"\ndrop_columns = ['age', \n 'title', \n 'rating',\n 'feedback_count', \n 'division',\n 'department',\n 'class']\n\n# + id=\"ld4s6GHIv9ZT\"\ndf_ml.drop(drop_columns, axis = 1, inplace = True)\n\n# + colab={\"base_uri\": \"https://localhost:8080/\"} id=\"1ePOIdvEv9ZT\" outputId=\"f23a105f-e0e6-44b6-bbc9-78b62d023f93\"\ndf_ml.info()\n\n# + id=\"Ncn7OfHGv9ZT\"\ndf_ml.rename(columns = {'review_text':'text', 'recommended':'recommend'}, inplace = True)\n\n# + colab={\"base_uri\": \"https://localhost:8080/\"} id=\"p5buBMPOv9ZU\" outputId=\"3ce10d8a-1e57-462d-f1bc-419f4a7b67a3\"\ndf_ml.columns\n\n# + [markdown] id=\"jJO0esyKv9ZU\"\n# ---\n# ---\n#\n\n# + [markdown] id=\"vl2h73vYCV38\"\n# ### Missing Value Detection\n\n# + colab={\"base_uri\": \"https://localhost:8080/\", \"height\": 80} id=\"q_wTPGCZv9ZU\" outputId=\"004458a1-1f68-4d76-bf68-1534de3b3d60\"\nmissing_values(df_ml)\n\n# + colab={\"base_uri\": \"https://localhost:8080/\", \"height\": 423} id=\"pn71H80Iv9ZU\" outputId=\"f2a877c1-f270-43c8-80a9-1fe52705a139\"\ndf_ml.isnull().melt(value_name=\"missing\")\n\n# + id=\"DpLw03eDv9ZV\"\nplt.figure(figsize = (10, 5))\n\nsns.displot(\n data = df_ml.isnull().melt(value_name = \"missing\"),\n y = \"variable\",\n hue = \"missing\",\n multiple = \"fill\",\n height = 9.25)\n\nplt.axvline(0.3, color = \"r\");\n\n# + id=\"EfMe0LNDv9ZV\"\ndf_ml = df_ml.dropna()\n\n# df_ml = df_ml.dropna(subset=['text'], axis=0)\n# df_ml = df_ml.reset_index(drop=True)\n\n# + colab={\"base_uri\": \"https://localhost:8080/\", \"height\": 49} id=\"MbPzrmhjv9ZV\" outputId=\"31104283-4e7a-4d8d-b61b-0c96698ec1e9\"\nmissing_values(df_ml)\n\n# + colab={\"base_uri\": \"https://localhost:8080/\"} id=\"vUA76-2lv9ZV\" outputId=\"52d64ee8-3166-4c45-9e20-678171606ad3\"\ndf_ml.info()\n\n# + colab={\"base_uri\": \"https://localhost:8080/\"} id=\"ZdxvRvHQv9ZW\" outputId=\"f9202493-2681-4307-c629-590893eabeba\"\ndf_ml[\"text\"].str.isspace().sum()\ndf_ml[df_ml[\"text\"].str.isspace() == True].index\n\n# + [markdown] id=\"CXvH8shBv9ZW\"\n# ---\n# ---\n#\n\n# + [markdown] id=\"WO_zLq2UCV39\"\n# ## 3. Text Mining\n#\n# Text is the most unstructured form of all the available data, therefore various types of noise are present in it. This means that the data is not readily analyzable without any pre-processing. The entire process of cleaning and standardization of text, making it noise-free and ready for analysis is known as **text preprocessing**.\n#\n# The three key steps of text preprocessing:\n#\n# - **Tokenization:**\n# This step is one of the top priorities when it comes to working on text mining. Tokenization is essentially splitting a phrase, sentence, paragraph, or an entire text document into smaller units, such as individual words or terms. Each of these smaller units are called tokens.\n#\n#\n# - **Noise Removal:**\n# Any piece of text which is not relevant to the context of the data and the end-output can be specified as the noise.\n# For example – language stopwords (commonly used words of a language – is, am, the, of, in etc), URLs or links, upper and lower case differentiation, punctuations and industry specific words. This step deals with removal of all types of noisy entities present in the text.\n#\n#\n# - **Lexicon Normalization:**\n# Another type of textual noise is about the multiple representations exhibited by single word.\n# For example – “play”, “player”, “played”, “plays” and “playing” are the different variations of the word – “play”. Though they mean different things, contextually they all are similar. This step converts all the disparities of a word into their normalized form (also known as lemma). \n# There are two methods of lexicon normalisation; **[Stemming or Lemmatization](https://www.guru99.com/stemming-lemmatization-python-nltk.html)**. Lemmatization is recommended for this case, because Lemmatization as this will return the root form of each word (rather than just stripping suffixes, which is stemming).\n#\n# As the first step change text to tokens and convertion all of the words to lower case. Next remove punctuation, bad characters, numbers and stop words. The second step is aimed to normalization them throught the Lemmatization method. \n\n# + [markdown] id=\"TPmXMWLSv9ZW\"\n# ### Tokenization, Noise Removal, Lexicon Normalization\n\n# + colab={\"base_uri\": \"https://localhost:8080/\", \"height\": 302} id=\"OS0NELtKv9ZX\" outputId=\"625f55d9-c44d-4a72-de9e-98a3764aede9\"\ndf_ml.head()\n\n# + id=\"IiADAGoryeca\"\n\n\n# + id=\"a4P8mq31v9ZX\"\nstop_words = stopwords.words('english')\n\n\n# + id=\"AHxNzAOiv9ZX\"\ndef cleaning(data):\n \n #1. Tokenize\n text_tokens = word_tokenize(data.replace(\"'\", \"\").lower()) \n \n #2. Remove Puncs\n tokens_without_punc = [w for w in text_tokens if w.isalpha()] \n \n #3. Removing Stopwords\n tokens_without_sw = [t for t in tokens_without_punc if t not in stop_words]\n \n #4. lemma\n text_cleaned = [WordNetLemmatizer().lemmatize(t) for t in tokens_without_sw]\n \n #joining\n return \" \".join(text_cleaned)\n\n\n# + id=\"K8GnwpJhv9ZX\"\ndf_ml[\"text\"] = df_ml[\"text\"].apply(cleaning)\ndf_ml[\"text\"].head()\n\n# + [markdown] id=\"bUMJUSwXv9ZY\"\n# ### Rare Words\n\n# + colab={\"base_uri\": \"https://localhost:8080/\"} id=\"Trt9eycEv9ZY\" outputId=\"4f20e647-2ec3-4b87-977c-0f45ff4cc302\"\n\" \".join(df_ml[\"text\"]).split()\n\n# + colab={\"base_uri\": \"https://localhost:8080/\"} id=\"cVG99X9kv9ZY\" outputId=\"0192b5de-ca5b-4e44-c9b1-4b02c7e6866c\"\nrare_words = pd.Series(\" \".join(df_ml[\"text\"]).split()).value_counts()\nrare_words\n\n# + id=\"V_cKUyRev9ZY\"\nrare_words = rare_words[rare_words <= 2] \n\n# + colab={\"base_uri\": \"https://localhost:8080/\"} id=\"iF-gLGE5v9ZY\" outputId=\"9aa24ee8-c38c-4988-d86f-9b69b56b2ce7\"\nrare_words.index\n\n# + colab={\"base_uri\": \"https://localhost:8080/\"} id=\"9mbwdiPav9ZY\" outputId=\"8f1db23e-bb71-4c6f-e24f-1e6ec8eae52b\"\ndf_ml[\"text\"] = df_ml[\"text\"].apply(lambda x: \" \".join([i for i in x.split() if i not in rare_words.index]))\ndf_ml[\"text\"].head()\n\n# + [markdown] id=\"_DTNRasaCV3-\"\n# ## 4. WordCloud - Repetition of Words\n#\n# Now we'll create a Word Clouds for reviews, representing most common words in each target class.\n#\n# Word Cloud is a data visualization technique used for representing text data in which the size of each word indicates its frequency or importance. Significant textual data points can be highlighted using a word cloud.\n#\n# We will create separate word clouds for positive and negative reviews. We can qualify a review as positive or negative, by looking at its recommended status.\n#\n# We can follow the steps below:\n#\n# - Detect Reviews\n# - Collect Words \n# - Create Word Cloud \n#\n\n# + [markdown] id=\"Dt11Bsv2v9ZZ\"\n# ### Detect Reviews (positive and negative separately)\n\n# + colab={\"base_uri\": \"https://localhost:8080/\", \"height\": 239} id=\"qmgm7ydAv9ZZ\" outputId=\"7b085a3a-3455-43c9-f271-776e5d79ccd0\"\ndf_ml[df_ml[\"recommend\"] == 0].head(3)\n\n# + colab={\"base_uri\": \"https://localhost:8080/\", \"height\": 204} id=\"oaMlDFNav9ZZ\" outputId=\"2c35a984-8a9f-45d0-d881-9562299bd753\"\ndf_ml[df_ml[\"recommend\"] == 1].head(3)\n\n# + [markdown] id=\"utYUiPPXv9ZZ\"\n# ### Collect Words (positive and negative separately)\n\n# + colab={\"base_uri\": \"https://localhost:8080/\"} id=\"Y3L4yJ10v9Za\" outputId=\"14d0bf4a-943c-4d53-8cad-a6d0855183c5\"\n\" \".join(df_ml[\"text\"]).split()\n\n# + colab={\"base_uri\": \"https://localhost:8080/\"} id=\"ZEMXpKBcv9Za\" outputId=\"1a9ce8b9-7a1e-4b4a-a46b-c0eb643100c8\"\npositive_words =\" \".join(df_ml[df_ml[\"recommend\"] == 1].text).split()\npositive_words\n\n# + colab={\"base_uri\": \"https://localhost:8080/\"} id=\"y25IC4r8v9Zb\" outputId=\"71a25161-079a-4c4f-cea3-eb009fb8eb55\"\nnegative_words = \" \".join(df_ml[df_ml[\"recommend\"] == 0].text).split()\nnegative_words \n\n# + colab={\"base_uri\": \"https://localhost:8080/\"} id=\"kQ24Wzh9v9Zb\" outputId=\"3f67acf7-af65-45c0-8741-1e6cb377c361\"\nlen(positive_words)\n\n# + colab={\"base_uri\": \"https://localhost:8080/\"} id=\"psMe52-1v9Zb\" outputId=\"2d39095b-a84b-43d1-bd75-defed12623c8\"\nlen(negative_words)\n\n# + [markdown] id=\"fFIETC8nv9Zc\"\n# ### Create Word Cloud (for most common words in recommended not recommended reviews separately)\n\n# + id=\"BN2gMoErv9Zc\"\nreview_text = df_ml[\"text\"]\n\n# + id=\"tn8yp6fiv9Zc\"\nall_words = \" \".join(review_text)\n\n# + colab={\"base_uri\": \"https://localhost:8080/\", \"height\": 35} id=\"C3mZvTAdv9Zd\" outputId=\"0fdf4f5b-2d4b-47ee-86a8-938299447a5a\"\nall_words[:100]\n\n# + colab={\"base_uri\": \"https://localhost:8080/\", \"height\": 394} id=\"62FzKxDBv9Zd\" outputId=\"ff01f9cb-3fe7-42a0-be79-0d153f12ea3a\"\nwordcloud = WordCloud(width = 800, height = 400, background_color = \"white\", max_words = 250).generate(all_words)\n\nplt.figure(figsize = (13, 13))\nplt.imshow(wordcloud)\nplt.axis(\"off\")\nplt.show()\n\n# + colab={\"base_uri\": \"https://localhost:8080/\", \"height\": 394} id=\"dkmPTspwv9Zd\" outputId=\"e2f8dcb0-ee11-4341-fa2f-89b77fe54b24\"\nwordcloud = WordCloud(width = 800, height = 400, background_color = \"white\", max_words = 250).generate(str(positive_words))\n\nplt.figure(figsize = (13, 13))\nplt.imshow(wordcloud)\nplt.axis(\"off\")\nplt.show()\n\n# + colab={\"base_uri\": \"https://localhost:8080/\", \"height\": 394} id=\"XpOk7Yp-v9Ze\" outputId=\"da564ca5-4724-4ac4-cb36-ebdc12c40c2e\"\nwordcloud = WordCloud(width = 800, height = 400, background_color = \"white\", max_words = 250).generate(str(negative_words))\n\nplt.figure(figsize = (13, 13))\nplt.imshow(wordcloud)\nplt.axis(\"off\")\nplt.show()\n\n# + [markdown] id=\"C01_6dLHv9Ze\"\n# ### Counting words\n\n# + id=\"0utNqUD9v9Ze\"\ncounter_all = Counter(word_tokenize(all_words))\ncounter_all.most_common(50)\n\n# + [markdown] id=\"-9_e1A26CV3_\"\n# ## 5. Sentiment Classification with Machine Learning and Deep Learning\n#\n# Before moving on to modeling, as data preprocessing steps we will need to perform **[vectorization](https://machinelearningmastery.com/prepare-text-data-machine-learning-scikit-learn/)** and **train-test split**. \n#\n# Machine learning algorithms most often take numeric feature vectors as input. Thus, when working with text documents, we need a way to convert each document into a numeric vector. This process is known as text vectorization. Commonly used vectorization approach that we will use here is to represent each text as a vector of word counts.\n#\n# At this moment, we have our review text column as a token (which has no punctuations and stopwords). We can use Scikit-learn’s CountVectorizer to convert the text collection into a matrix of token counts. We can imagine this resulting matrix as a 2-D matrix, where each row is a unique word, and each column is a review.\n#\n# After performing data preprocessing, we will build our models using following classification algorithms:\n#\n# - Logistic Regression,\n# - Naive Bayes,\n# - Support Vector Machine,\n# - Random Forest,\n# - Ada Boosting\n# - Deep Learning Model.\n\n# + [markdown] id=\"iGH_HL4mv9Zf\"\n# ### Train - Test Split\n\n# + [markdown] id=\"T2SMUUrxCV3_\"\n# To run machine learning algorithms we need to convert text files into numerical feature vectors. We will use bag of words model for our analysis.\n#\n# First we spliting the data into train and test sets:\n\n# + id=\"8pUAhpphv9Zf\"\nX = df_ml[\"text\"].values\ny = df_ml[\"recommend\"].map({0:1, 1:0}).values\n\n# + id=\"YAKxwBPav9Zf\"\nfrom sklearn.model_selection import train_test_split\n\nX_train, X_test, y_train, y_test = train_test_split(X, y, test_size = 0.2, stratify = y, random_state = 101)\n\n# + [markdown] id=\"adhWeL_iCV3_\"\n# In the next step we create a numerical feature vector for each document:\n\n# + [markdown] id=\"jEG4zGRXv9Zg\"\n# ### Count Vectorization\n\n# + id=\"o5gif1Auv9Zg\"\nfrom sklearn.feature_extraction.text import CountVectorizer\n\n# + id=\"3u1Kaw9pv9Zg\"\nvectorizer = CountVectorizer()\n\nX_train_count = vectorizer.fit_transform(X_train)\nX_test_count = vectorizer.transform(X_test)\n\n# + id=\"ymOWlLycv9Zg\"\nX_train_count\n\n# + id=\"pkeRYsjuv9Zh\"\nX_test_count\n\n# + id=\"gsbgqVkRv9Zh\"\nlen(X_train_count.toarray())\n\n# + id=\"yw3QTbOUv9Zh\"\nX_train_count.toarray()\n\n# + id=\"OHw3TrTYv9Zh\"\nlen(X_test_count.toarray())\n\n# + id=\"GFnYrtAfv9Zi\"\nX_test_count.toarray()\n\n# + id=\"wCCHTF9hv9Zi\"\npd.DataFrame(X_train_count.toarray(), columns = vectorizer.get_feature_names())\n\n# + id=\"5pfQWdrSv9Zi\"\nX_train\n\n# + [markdown] id=\"luSB2bzYv9Zi\"\n# ### TF-IDF\n\n# + id=\"UupPU_1Mv9Zi\"\nfrom sklearn.feature_extraction.text import TfidfVectorizer\n\n# + id=\"XMnRd3Thv9Zi\"\ntf_idf_vectorizer = TfidfVectorizer()\n\nX_train_tf_idf = tf_idf_vectorizer.fit_transform(X_train)\nX_test_tf_idf = tf_idf_vectorizer.transform(X_test)\n\n# + id=\"LPHRAfUQv9Zj\"\nX_train_tf_idf.toarray()\n\n# + id=\"LQCcZ_C2v9Zj\"\npd.DataFrame(X_train_tf_idf.toarray(), columns = tf_idf_vectorizer.get_feature_names())\n\n# + id=\"lBj4FzKMv9Zj\"\n\n\n# + [markdown] id=\"jp58tfdWv9Zj\"\n# ## Modelling with Machine Learning Models\n\n# + id=\"o_WXxDm-v9Zj\"\nfrom sklearn.metrics import plot_confusion_matrix, confusion_matrix, classification_report, accuracy_score, f1_score, recall_score, precision_score, average_precision_score\n\n\n# + id=\"mmPD-z6uv9Zk\"\ndef eval(model, X_train, X_test):\n y_pred = model.predict(X_test)\n y_pred_train = model.predict(X_train)\n \n print(confusion_matrix(y_test, y_pred))\n print(\"Test_Set\")\n print(classification_report(y_test,y_pred))\n print(\"Train_Set\")\n print(classification_report(y_train,y_pred_train))\n plot_confusion_matrix(model, X_test, y_test, cmap=\"plasma\")\n\n\n# + [markdown] id=\"xaVrkxRpCV3_\"\n# ## Logistic Regression\n\n# + [markdown] id=\"baRfxJF9v9Zk\"\n# ### CountVectorizer \n\n# + id=\"qdcBeWscv9Zl\"\nfrom sklearn.linear_model import LogisticRegression\n\nlogreg_count = LogisticRegression(C = 0.1, max_iter = 1000, class_weight = 'balanced', random_state = 101)\nlogreg_count.fit(X_train_count,y_train)\n\n# + id=\"o22BMB6ev9Zl\"\nprint(\"LogReg_Count Model\")\nprint (\"------------------\")\neval(logreg_count, X_train_count, X_test_count)\n\n# + id=\"cIU6QzbZv9Zl\"\nimport random\nimport pylab as pl\nfrom sklearn.metrics import precision_recall_curve\nfrom sklearn.metrics import auc\n\n# + id=\"NvpS9Qjxv9Zm\"\nfrom yellowbrick.classifier import PrecisionRecallCurve\n\nviz = PrecisionRecallCurve(\n LogisticRegression(C = 0.1, max_iter = 1000, class_weight= \"balanced\", random_state=101),\n classes = logreg_count.classes_,\n per_class = True,\n cmap = \"Set1\")\n\nfig, ax = plt.subplots(figsize = (10, 6))\nax.set_facecolor('#eafff5')\n\nviz.fit(X_train_count,y_train)\nviz.score(X_test_count, y_test)\nviz.show();\n\n# + id=\"x1jmUea_v9Zm\"\ny_pred = logreg_count.predict(X_test_count)\nlog_count_rec = recall_score(y_test, y_pred, pos_label = 0, average = None)\nlog_count_f1 = f1_score(y_test, y_pred, pos_label = 0, average = None)\nlog_AP = viz.score_\n\n# + id=\"ddVRexdAv9Zm\"\nprint(\"viz.score_ : \", viz.score_)\nprint(\"LogReg_Count_rec : \", log_count_rec)\nprint(\"LogReg_Count_f1 : \", log_count_f1)\nprint(\"LogReg_Count_AP : \", log_AP)\n\n# + [markdown] id=\"XknI1yAov9Zn\"\n# ### CountVectorizer ***Cross Validation***\n\n# + id=\"OLlv8btvv9Zn\"\nfrom sklearn.metrics import make_scorer\nfrom sklearn.model_selection import cross_val_score\n\ncustom_scorer = {'accuracy': make_scorer(accuracy_score),\n 'precision-0': make_scorer(precision_score, pos_label = 0),\n 'recall-0': make_scorer(recall_score, pos_label = 0),\n 'f1-0': make_scorer(f1_score, pos_label = 0),\n 'precision-1': make_scorer(precision_score, pos_label = 1),\n 'recall-1': make_scorer(recall_score, pos_label = 1),\n 'f1-1': make_scorer(f1_score, pos_label = 1)\n }\n\nfor i, j in custom_scorer.items():\n model = LogisticRegression(C = 0.1, max_iter = 1000, class_weight = \"balanced\", random_state = 101)\n scores = cross_val_score(model, X_train_count, y_train, cv = 10, scoring = j).mean()\n if i == \"recall-0\":\n log_count_rec = scores\n elif i == \"f1-0\":\n log_count_f1 = scores\n print(f\" {i:20} score for LogReg_Count : {scores}\\n\")\n\n# + id=\"GIkMy9tQv9Zn\"\nimport random\nimport pylab as pl\nfrom sklearn.metrics import precision_recall_curve\nfrom sklearn.metrics import auc\n\n# + id=\"0l0XSj0wv9Zn\"\nfrom yellowbrick.classifier import PrecisionRecallCurve\nviz = PrecisionRecallCurve(\n LogisticRegression(C = 0.1, max_iter = 1000, class_weight = \"balanced\", random_state = 101),\n classes = logreg_count.classes_,\n per_class = True,\n cmap = \"Set1\")\n\nfig, ax = plt.subplots(figsize = (10, 6))\nax.set_facecolor('#eafff5')\n\nviz.fit(X_train_count,y_train)\nviz.score(X_test_count, y_test)\nviz.show();\n\n# + id=\"KAjOgqZTv9Zo\"\nLogReg_Count_AP = viz.score_\nLogReg_Count_AP\n\n# + [markdown] id=\"a3DlymQlv9Zo\"\n# ### TF-IDF \n\n# + id=\"zerDb_t1v9Zo\"\nlogreg_tfidf = LogisticRegression(C = 1, max_iter = 1000, class_weight = \"balanced\", random_state = 101)\nlogreg_tfidf.fit(X_train_tf_idf,y_train)\n\n# + id=\"-CzzYDuJv9Zp\"\nprint(\"LogReg_TFIDF Model\")\nprint(\"------------------\")\neval(logreg_tfidf, X_train_tf_idf, X_test_tf_idf)\n\n# + id=\"bfxgW3W2v9Zp\"\nfrom yellowbrick.classifier import PrecisionRecallCurve\nviz = PrecisionRecallCurve(\n LogisticRegression(C = 1, max_iter = 1000, class_weight = \"balanced\", random_state=101),\n classes = logreg_count.classes_,\n per_class = True,\n cmap = \"Set1\")\n\nfig, ax = plt.subplots(figsize = (10, 6))\nax.set_facecolor('#eafff5')\n\nviz.fit(X_train_tf_idf,y_train)\nviz.score(X_test_tf_idf, y_test)\nviz.show();\n\n# + id=\"0TgitnNav9Zp\"\ny_pred = logreg_tfidf.predict(X_test_tf_idf)\nlog_tf_idf_rec = recall_score(y_test, y_pred, pos_label = 0, average = None)\nlog_tf_idf_f1 = f1_score(y_test, y_pred, pos_label = 0, average = None)\nlog_tf_idf_AP = viz.score_\n\n# + id=\"t_MGGBnrv9Zp\"\nprint(\"viz.score_ : \", viz.score_)\nprint(\"LogReg_TFIDF_rec : \", log_tf_idf_rec)\nprint(\"LogReg_TFIDF_f1 : \", log_tf_idf_f1)\nprint(\"LogReg_TFIDF_AP : \", log_tf_idf_AP)\n\n# + [markdown] id=\"0dZHFItHv9Zq\"\n# ### TF-IDF ***Cross Validation***\n\n# + id=\"mLtEXE4Jv9Zq\"\ncustom_scorer = {'accuracy': make_scorer(accuracy_score),\n 'precision-0': make_scorer(precision_score, pos_label = 0),\n 'recall-0': make_scorer(recall_score, pos_label = 0),\n 'f1-0': make_scorer(f1_score, pos_label = 0),\n 'precision-1': make_scorer(precision_score, pos_label = 1),\n 'recall-1': make_scorer(recall_score, pos_label = 1),\n 'f1-1': make_scorer(f1_score, pos_label = 1)\n }\n\nfor i, j in custom_scorer.items():\n LogisticRegression(C = 1, max_iter = 1000, random_state = 101, class_weight = \"balanced\")\n scores = cross_val_score(model, X_train_tf_idf, y_train, cv = 10, scoring = j).mean()\n if i == \"recall-1\":\n log_tfidf_rec = scores\n elif i == \"f1-1\":\n log_tfidf_f1 = scores\n print(f\" {i:20} score for LogReg_TFIDF : {scores}\\n\")\n\n# + id=\"pfDWrLmkv9Zq\"\nviz = PrecisionRecallCurve(\n LogisticRegression(C = 1, max_iter = 1000, random_state = 101, class_weight = \"balanced\"),\n classes = logreg_tfidf.classes_,\n per_class = True,\n cmap = \"Set1\")\n\nfig, ax = plt.subplots(figsize = (10, 6))\nax.set_facecolor('#eafff5')\n\nviz.fit(X_train_tf_idf,y_train)\nviz.score(X_test_tf_idf, y_test)\nviz.show();\n\n# + id=\"Wplrh-kRv9Zr\"\nLogReg_TFIDF_AP = viz.score_\nLogReg_TFIDF_AP\n\n# + [markdown] id=\"QACRU9G3CV4A\"\n# ## Naive Bayes \n#\n# ### Countvectorizer ***MultinomialNB***\n\n# + id=\"sWT0VKs8v9Zr\"\nfrom sklearn.naive_bayes import MultinomialNB\n\n# + id=\"AmNvyXmkv9Zr\"\nnbmulti_count = MultinomialNB()\nnbmulti_count.fit(X_train_count,y_train)\n\n# + id=\"TxbYlnkLv9Zr\"\nprint(\"NBMulti_Count Model\")\nprint(\"-------------------\")\neval(nbmulti_count, X_train_count, X_test_count)\n\n# + id=\"8Zwn21Z0v9Zs\"\nfrom yellowbrick.classifier import PrecisionRecallCurve\n\nviz = PrecisionRecallCurve(\n MultinomialNB(),\n classes = nbmulti_count.classes_, \n per_class = True,\n cmap = \"Set1\")\n\nfig, ax = plt.subplots(figsize = (10, 6))\nax.set_facecolor('#eafff5')\n\nviz.fit(X_train_count,y_train)\nviz.score(X_test_count, y_test)\nviz.show();\n\n# + id=\"3LAmexMMv9Zs\"\ny_pred = nbmulti_count.predict(X_test_count)\nnb_multi_count_rec = recall_score(y_test, y_pred, pos_label = 0, average = None)\nnb_multi_count_f1 = f1_score(y_test, y_pred, pos_label = 0, average = None)\nnb_multi_count_AP = viz.score_\n\n# + id=\"yVOR3axqv9Zs\"\nprint(\"viz.score_ : \", viz.score_)\nprint(\"NBMulti_Count_rec : \", nb_multi_count_rec)\nprint(\"NBMulti_Count_f1 : \", nb_multi_count_f1)\nprint(\"NBMulti_Count_AP : \", nb_multi_count_AP)\n\n# + [markdown] id=\"u9omyYilv9Zs\"\n# ### Countvectorize ***MultinomialNB with Cross Validation***\n\n# + id=\"n7CG7XIQv9Zt\"\ncustom_scorer = {'accuracy': make_scorer(accuracy_score),\n 'precision-0': make_scorer(precision_score, pos_label=0),\n 'recall-0': make_scorer(recall_score, pos_label=0),\n 'f1-0': make_scorer(f1_score, pos_label=0),\n 'precision-1': make_scorer(precision_score, pos_label=1),\n 'recall-1': make_scorer(recall_score, pos_label=1),\n 'f1-1': make_scorer(f1_score, pos_label=1)\n }\n\nfor i, j in custom_scorer.items():\n model = MultinomialNB()\n scores = cross_val_score(model, X_train_count, y_train, cv = 10, scoring = j).mean()\n if i == \"recall-1\":\n nbm_count_rec = scores\n elif i == \"f1-1\":\n nbm_count_f1 = scores\n print(f\" {i:20} score for NBMulti_Count : {scores}\\n\")\n\n# + id=\"ldwovz3uv9Zt\"\nfrom yellowbrick.classifier import PrecisionRecallCurve\n\nviz = PrecisionRecallCurve(\n MultinomialNB(),\n classes = nbmulti_count.classes_, \n per_class = True,\n cmap = \"Set1\"\n)\n\nfig, ax = plt.subplots(figsize = (10, 6))\nax.set_facecolor('#eafff5')\n\nviz.fit(X_train_count,y_train)\nviz.score(X_test_count, y_test)\nviz.show();\n\n# + id=\"_z1bDfz7v9Zt\"\nNBMulti_Count_AP = viz.score_\nNBMulti_Count_AP\n\n# + [markdown] id=\"lgv7xLkhv9Zt\"\n# ### Countvectorize ***BernoulliNB***\n\n# + id=\"c1byMA4pv9Zt\"\nfrom sklearn.naive_bayes import BernoulliNB\n\n# + id=\"t62dx5-jv9Zu\"\nnbberno_count = BernoulliNB()\nnbberno_count.fit(X_train_count,y_train)\n\n# + id=\"-F2s8ZOtv9Zu\"\nprint(\"NBBerno_Count Model\")\nprint(\"-------------------\")\neval(nbberno_count, X_train_count, X_test_count)\n\n# + id=\"scY4_ExNv9Zu\"\nfrom yellowbrick.classifier import PrecisionRecallCurve\n\nviz = PrecisionRecallCurve(\n BernoulliNB(),\n classes = nbberno_count.classes_, \n per_class = True,\n cmap = \"Set1\")\n\nfig, ax = plt.subplots(figsize = (10, 6))\nax.set_facecolor('#eafff5')\n\nviz.fit(X_train_count,y_train)\nviz.score(X_test_count, y_test)\nviz.show();\n\n# + id=\"b3Uf0aMNv9Zu\"\ny_pred = nbberno_count.predict(X_test_count)\nnb_ber_count_rec = recall_score(y_test, y_pred, pos_label = 0, average = None)\nnb_ber_count_f1 = f1_score(y_test, y_pred, pos_label = 0, average = None)\nnb_ber_count_AP = viz.score_\n\n# + id=\"ruNDtYO5v9Zv\"\nprint(\"viz.score_ : \", viz.score_)\nprint(\"NBBerno_Count_rec : \", nb_ber_count_rec)\nprint(\"NBBerno_Count_f1 : \", nb_ber_count_f1)\nprint(\"NBBerno_Count_AP : \", nb_ber_count_AP)\n\n# + [markdown] id=\"WMeIqqUfv9Zv\"\n# ### Countvectorize ***BernoulliNB with Cross Validation***\n\n# + id=\"GNkAogtGv9Zv\"\ncustom_scorer = {'accuracy': make_scorer(accuracy_score),\n 'precision-0': make_scorer(precision_score, pos_label = 0),\n 'recall-0': make_scorer(recall_score, pos_label = 0),\n 'f1-0': make_scorer(f1_score, pos_label = 0),\n 'precision-1': make_scorer(precision_score, pos_label = 1),\n 'recall-1': make_scorer(recall_score, pos_label = 1),\n 'f1-1': make_scorer(f1_score, pos_label = 1)\n }\n\nfor i, j in custom_scorer.items():\n model = BernoulliNB()\n scores = cross_val_score(model, X_train_count, y_train, cv = 10, scoring = j).mean()\n if i == \"recall-1\":\n nbb_count_rec = scores\n elif i == \"f1-1\":\n nbb_count_f1 = scores\n print(f\" {i:20} score for NBBerno_Count : {scores}\\n\")\n\n# + id=\"lgXQCq5Xv9Zv\"\nfrom yellowbrick.classifier import PrecisionRecallCurve\n\nviz = PrecisionRecallCurve(\n BernoulliNB(),\n classes = nbberno_count.classes_, \n per_class = True,\n cmap = \"Set1\")\n\nfig, ax = plt.subplots(figsize = (10, 6))\nax.set_facecolor('#eafff5')\n\nviz.fit(X_train_count,y_train)\nviz.score(X_test_count, y_test)\nviz.show();\n\n# + id=\"s0APEpBLv9Zw\"\nNBBerno_Count_AP = viz.score_\nNBBerno_Count_AP\n\n# + [markdown] id=\"y4ueT2w9v9Zw\"\n# ### TF-IDF ***MultinomialNB***\n\n# + id=\"QrujBXI8v9Zw\"\nnbmulti_tfidf = MultinomialNB()\nnbmulti_tfidf.fit(X_train_tf_idf,y_train)\n\n# + id=\"1iivjyiXv9Zw\"\nprint(\"NBMulti_TFIDF MODEL\")\nprint(\"-------------------\")\neval(nbmulti_tfidf, X_train_tf_idf, X_test_tf_idf)\n\n# + id=\"wxvjkPGfv9Zx\"\nfrom yellowbrick.classifier import PrecisionRecallCurve\n\nviz = PrecisionRecallCurve(\n MultinomialNB(),\n classes = nbmulti_tfidf.classes_,\n per_class = True,\n cmap = \"Set1\")\n\nfig, ax = plt.subplots(figsize = (10, 6))\nax.set_facecolor('#eafff5')\n\nviz.fit(X_train_tf_idf,y_train)\nviz.score(X_test_tf_idf, y_test)\nviz.show();\n\n# + id=\"Wao1DZwBv9Zx\"\ny_pred = nbmulti_tfidf.predict(X_test_tf_idf)\nnb_multi_tf_idf_rec = recall_score(y_test, y_pred, pos_label = 0, average = None)\nnb_multi_tf_idf_f1 = f1_score(y_test, y_pred, pos_label = 0, average = None)\nnb_multi_tf_idf_AP = viz.score_\n\n# + id=\"TB3a_juLv9Zy\"\nprint(\"viz.score_ : \", viz.score_)\nprint(\"NBMulti_TFIDF_rec : \", nb_multi_tf_idf_rec)\nprint(\"NBMulti_TFIDF_f1 : \", nb_multi_tf_idf_f1)\nprint(\"NBMulti_TFIDF_AP : \", nb_multi_tf_idf_AP)\n\n# + [markdown] id=\"mI4z5wHKv9Zy\"\n# ### TF-IDF ***MultinomialNB with Cross Validation***\n\n# + id=\"iUw1DcM_v9Zz\"\ncustom_scorer = {'accuracy': make_scorer(accuracy_score),\n 'precision-0': make_scorer(precision_score, pos_label = 0),\n 'recall-0': make_scorer(recall_score, pos_label = 0),\n 'f1-0': make_scorer(f1_score, pos_label = 0),\n 'precision-1': make_scorer(precision_score, pos_label = 1),\n 'recall-1': make_scorer(recall_score, pos_label = 1),\n 'f1-1': make_scorer(f1_score, pos_label = 1)\n }\n\nfor i, j in custom_scorer.items():\n model = MultinomialNB()\n scores = cross_val_score(model, X_train_tf_idf, y_train, cv = 10, scoring = j).mean()\n if i == \"recall-1\":\n nbm_tfidf_rec = scores\n elif i == \"f1-1\":\n nbm_tfidf_f1 = scores\n print(f\" {i:20} score for NBMulti_TFIDF : {scores}\\n\")\n\n# + id=\"fdBZT6VAv9Zz\"\nfrom yellowbrick.classifier import PrecisionRecallCurve\n\nviz = PrecisionRecallCurve(\n MultinomialNB(),\n classes = nbmulti_tfidf.classes_,\n per_class = True,\n cmap = \"Set1\")\n\nfig, ax = plt.subplots(figsize = (10, 6))\nax.set_facecolor('#eafff5')\n\nviz.fit(X_train_tf_idf,y_train)\nviz.score(X_test_tf_idf, y_test)\nviz.show();\n\n# + id=\"g8q8OXRJv9Zz\"\nNBMulti_TFIDF_AP = viz.score_\nNBMulti_TFIDF_AP\n\n# + [markdown] id=\"oZ-6eb1Hv9Z0\"\n# ### TF-IDF ***BernoulliNB***\n\n# + id=\"uWp7MCOSv9Z0\"\nnbberno_tfidf = BernoulliNB()\nnbberno_tfidf.fit(X_train_tf_idf,y_train)\n\n# + id=\"YBNl8_7Hv9Z0\"\nprint(\"NBBerno_TFIDF MODEL\")\nprint(\"-------------------\")\neval(nbberno_tfidf, X_train_tf_idf, X_test_tf_idf)\n\n# + id=\"ZqNmLStbv9Z0\"\nfrom yellowbrick.classifier import PrecisionRecallCurve\n\nviz = PrecisionRecallCurve(\n BernoulliNB(),\n classes = nbberno_tfidf.classes_, \n per_class = True,\n cmap = \"Set1\")\n\nfig, ax = plt.subplots(figsize = (10, 6))\nax.set_facecolor('#eafff5')\n\nviz.fit(X_train_tf_idf,y_train)\nviz.score(X_test_tf_idf, y_test)\nviz.show();\n\n# + id=\"U5LrZuZYv9Z1\"\ny_pred = nbberno_tfidf.predict(X_test_tf_idf)\nnb_ber_tf_idf_rec = recall_score(y_test, y_pred, pos_label = 0, average = None)\nnb_ber_tf_idf_f1 = f1_score(y_test, y_pred, pos_label = 0, average = None)\nnb_ber_tf_idf_AP = viz.score_\n\n# + id=\"YDb14KNxv9Z1\"\nprint(\"viz.score_ : \", viz.score_)\nprint(\"NBBerno_TFIDF_rec : \", nb_multi_tf_idf_rec)\nprint(\"NBBerno_TFIDF_f1 : \", nb_multi_tf_idf_f1)\nprint(\"NBBerno_TFIDF_AP : \", nb_multi_tf_idf_AP)\n\n# + [markdown] id=\"GBNb3nJJv9Z1\"\n# ### TF-IDF ***BernoulliNB with Cross Validation***\n\n# + id=\"zPH4sxB_v9Z1\"\ncustom_scorer = {'accuracy': make_scorer(accuracy_score),\n 'precision-0': make_scorer(precision_score, pos_label = 0),\n 'recall-0': make_scorer(recall_score, pos_label = 0),\n 'f1-0': make_scorer(f1_score, pos_label = 0),\n 'precision-1': make_scorer(precision_score, pos_label = 1),\n 'recall-1': make_scorer(recall_score, pos_label = 1),\n 'f1-1': make_scorer(f1_score, pos_label = 1)\n }\n\nfor i, j in custom_scorer.items():\n model = BernoulliNB()\n scores = cross_val_score(model, X_train_tf_idf, y_train, cv = 10, scoring = j).mean()\n if i == \"recall-1\":\n nbb_tfidf_rec = scores\n elif i == \"f1-1\":\n nbb_tfidf_f1 = scores\n print(f\" {i:20} score for NBBerno_TFIDF : {scores}\\n\")\n\n# + id=\"jKEVYR8Sv9Z2\"\nfrom yellowbrick.classifier import PrecisionRecallCurve\n\nviz = PrecisionRecallCurve(\n BernoulliNB(),\n classes = nbberno_tfidf.classes_, \n per_class = True,\n cmap = \"Set1\")\n\nfig, ax = plt.subplots(figsize = (10, 6))\nax.set_facecolor('#eafff5')\n\nviz.fit(X_train_tf_idf,y_train)\nviz.score(X_test_tf_idf, y_test)\nviz.show();\n\n# + id=\"mBI43OmKv9Z2\"\nNBBerno_TFIDF_AP = viz.score_\nNBBerno_TFIDF_AP\n\n# + [markdown] id=\"OSkbnJJiCV4A\"\n# ## Support Vector Machine (SVM)\n#\n# ### Countvectorizer\n\n# + id=\"2wT6KoqDv9Z2\"\nfrom sklearn.svm import LinearSVC\nsvc_count = LinearSVC(C = 0.01, class_weight = \"balanced\", random_state = 101) \nsvc_count.fit(X_train_count,y_train)\n\n# + id=\"57S7S6GVv9Z2\"\nprint(\"SVC_Count Model\")\nprint(\"---------------\")\neval(svc_count, X_train_count, X_test_count)\n\n# + id=\"z1J_nfVHv9Z3\"\nviz = PrecisionRecallCurve(\n LinearSVC(C = 0.01, class_weight = \"balanced\", random_state = 101),\n classes = svc_count.classes_,\n per_class = True,\n cmap = \"Set1\")\n\nfig, ax = plt.subplots(figsize = (10, 6))\nax.set_facecolor('#eafff5')\n\nviz.fit(X_train_count,y_train)\nviz.score(X_test_count, y_test)\nviz.show();\n\n# + id=\"ilq5U2yRv9Z3\"\ny_pred = svc_count.predict(X_test_count)\nsvc_count_rec = recall_score(y_test, y_pred, pos_label = 0, average = None)\nsvc_count_f1 = f1_score(y_test, y_pred, pos_label = 0, average = None)\nsvc_count_AP = viz.score_\n\n# + id=\"KLZMoM0ov9Z3\"\nprint(\"viz.score_ : \", viz.score_)\nprint(\"SVC_Count_rec : \", svc_count_rec)\nprint(\"SVC_Count_f1 : \", svc_count_f1)\nprint(\"SVC_Count_AP : \", svc_count_AP)\n\n# + [markdown] id=\"xOmM8aUkv9Z3\"\n# ### CountVectorizer ***With Cross Validation***\n\n# + id=\"5VmR9zSvv9Z3\"\ncustom_scorer = {'accuracy': make_scorer(accuracy_score),\n 'precision-0': make_scorer(precision_score, pos_label = 0),\n 'recall-0': make_scorer(recall_score, pos_label = 0),\n 'f1-0': make_scorer(f1_score, pos_label = 0),\n 'precision-1': make_scorer(precision_score, pos_label = 1),\n 'recall-1': make_scorer(recall_score, pos_label = 1),\n 'f1-1': make_scorer(f1_score, pos_label = 1)\n }\n\nfor i, j in custom_scorer.items():\n model = LinearSVC(C = 0.01, class_weight = \"balanced\", random_state = 101)\n scores = cross_val_score(model, X_train_count, y_train, cv = 10, scoring = j).mean()\n if i == \"recall-1\":\n svc_count_rec = scores\n elif i == \"f1-1\":\n svc_count_f1 = scores\n print(f\" {i:20} score for SVC_Count : {scores}\\n\")\n\n# + id=\"z1MXuCZZv9Z4\"\nviz = PrecisionRecallCurve(\n LinearSVC(C = 0.01, class_weight = \"balanced\", random_state = 101),\n classes = svc_count.classes_,\n per_class = True,\n cmap = \"Set1\")\n\nfig, ax = plt.subplots(figsize = (10, 6))\nax.set_facecolor('#eafff5')\n\nviz.fit(X_train_count,y_train)\nviz.score(X_test_count, y_test)\nviz.show();\n\n# + id=\"4LAuy3eAv9Z4\"\nSVC_Count_AP = viz.score_\nSVC_Count_AP\n\n# + [markdown] id=\"p-lC6Mbyv9Z4\"\n# ### TD-IDF\n\n# + id=\"UsfkaspTv9Z4\"\nsvc_tf_idf = LinearSVC(C = 0.01, class_weight = \"balanced\", random_state = 101) \nsvc_tf_idf.fit(X_train_tf_idf,y_train)\n\n# + id=\"xbMkBpzDv9Z4\"\nprint(\"SVC_TFIDF Model\")\nprint(\"---------------\")\neval(svc_tf_idf, X_train_tf_idf, X_test_tf_idf)\n\n# + id=\"ZWILabobv9Z5\"\nviz = PrecisionRecallCurve(\n LinearSVC(C = 0.01, class_weight = \"balanced\", random_state = 101),\n classes = svc_tf_idf.classes_,\n per_class = True,\n cmap = \"Set1\"\n)\n\nfig, ax = plt.subplots(figsize = (10, 6))\nax.set_facecolor('#eafff5')\n\nviz.fit(X_train_tf_idf,y_train)\nviz.score(X_test_tf_idf, y_test)\nviz.show();\n\n# + id=\"uwuMHECvv9Z5\"\ny_pred = svc_tf_idf.predict(X_test_tf_idf)\nsvc_tf_idf_rec = recall_score(y_test, y_pred, pos_label = 0, average = None)\nsvc_tf_idf_f1 = f1_score(y_test, y_pred, pos_label = 0, average = None)\nsvc_tf_idf_AP = viz.score_\n\n# + id=\"rBN1bYlQv9Z5\"\nprint(\"viz.score_ : \", viz.score_)\nprint(\"SVC_TFIDF_rec : \", svc_tf_idf_rec)\nprint(\"SVC_TFIDF_f1 : \", svc_tf_idf_f1)\nprint(\"SVC_TFIDF_AP : \", svc_tf_idf_AP)\n\n# + [markdown] id=\"QKfWPfPHv9Z5\"\n# ### TFIDF ***With Cross Validation***\n\n# + id=\"0_9xfBshv9Z5\"\ncustom_scorer = {'accuracy': make_scorer(accuracy_score),\n 'precision-0': make_scorer(precision_score, pos_label = 0),\n 'recall-0': make_scorer(recall_score, pos_label = 0),\n 'f1-0': make_scorer(f1_score, pos_label = 0),\n 'precision-1': make_scorer(precision_score, pos_label = 1),\n 'recall-1': make_scorer(recall_score, pos_label = 1),\n 'f1-1': make_scorer(f1_score, pos_label = 1)\n }\n\nfor i, j in custom_scorer.items():\n model = LinearSVC(C = 0.01, class_weight = \"balanced\", random_state = 101)\n scores = cross_val_score(model, X_train_tf_idf, y_train, cv = 10, scoring = j).mean()\n if i == \"recall-1\":\n svc_tfidf_rec = scores\n elif i == \"f1-1\":\n svc_tfidf_f1 = scores\n print(f\" {i:20} score for SVC_TFIDF : {scores}\\n\")\n\n# + id=\"tu8xeDShv9Z6\"\nviz = PrecisionRecallCurve(\n LinearSVC(C = 0.01, class_weight = \"balanced\", random_state = 101),\n classes = svc_tf_idf.classes_,\n per_class = True,\n cmap = \"Set1\")\n\nfig, ax = plt.subplots(figsize=(10, 6))\nax.set_facecolor('#eafff5')\n\nviz.fit(X_train_tf_idf,y_train)\nviz.score(X_test_tf_idf, y_test)\nviz.show();\n\n# + id=\"Mtx_KduXv9Z6\"\nSVC_TFIDF_AP = viz.score_\nSVC_TFIDF_AP\n\n# + [markdown] id=\"qTECEchfCV4A\"\n# ## Random Forest\n#\n# ### CountVectorizer\n\n# + id=\"X8594rsIv9Z6\"\nfrom sklearn.ensemble import RandomForestClassifier\n\nrf_count = RandomForestClassifier(n_estimators = 200, max_depth = 11, class_weight = \"balanced\", random_state = 101, n_jobs = -1)\nrf_count.fit(X_train_count, y_train)\n\n# + id=\"laLglHa4v9Z7\"\nprint(\"RF_Count Model\")\nprint(\"--------------\")\neval(rf_count, X_train_count, X_test_count)\n\n# + id=\"9sPsAz_0v9Z7\"\nviz = PrecisionRecallCurve(\n RandomForestClassifier(n_estimators = 200, max_depth = 11, class_weight = \"balanced\", random_state = 101, n_jobs = -1),\n classes = rf_count.classes_,\n per_class = True,\n cmap = \"Set1\")\n\nfig, ax = plt.subplots(figsize=(10, 6))\nax.set_facecolor('#eafff5')\n\nviz.fit(X_train_count,y_train)\nviz.score(X_test_count, y_test)\nviz.show();\n\n# + id=\"fBtweNvnv9Z7\"\ny_pred = rf_count.predict(X_test_count)\nrf_count_rec = recall_score(y_test, y_pred, pos_label = 0, average = None)\nrf_count_f1 = f1_score(y_test, y_pred, pos_label = 0, average = None)\nrf_count_AP = viz.score_\n\n# + id=\"Bg4BZx0fv9Z7\"\nprint(\"viz.score_ : \", viz.score_)\nprint(\"RF_Count_rec : \", rf_count_rec)\nprint(\"RF_Count_f1 : \", rf_count_f1)\nprint(\"RF_Count_AP : \", rf_count_AP)\n\n# + [markdown] id=\"R6CCCsAKv9Z7\"\n# ### CountVectorizer with Cross Validation\n\n# + id=\"nVFBIr2Gv9Z8\"\ncustom_scorer = {'accuracy': make_scorer(accuracy_score),\n 'precision-0': make_scorer(precision_score, pos_label = 0),\n 'recall-0': make_scorer(recall_score, pos_label = 0),\n 'f1-0': make_scorer(f1_score, pos_label = 0),\n 'precision-1': make_scorer(precision_score, pos_label = 1),\n 'recall-1': make_scorer(recall_score, pos_label = 1),\n 'f1-1': make_scorer(f1_score, pos_label = 1)\n }\n\nfor i, j in custom_scorer.items():\n model = RandomForestClassifier(n_estimators = 200, max_depth = 11, class_weight = \"balanced\", random_state = 101, n_jobs = -1)\n scores = cross_val_score(model, X_train_count, y_train, cv = 10, scoring = j).mean()\n if i == \"recall-1\":\n rf_count_rec = scores\n elif i == \"f1-1\":\n rf_count_f1 = scores\n print(f\" {i:20} score for RF_Count : {scores}\\n\")\n\n# + id=\"soJy4dFQv9Z8\"\nviz = PrecisionRecallCurve(\n RandomForestClassifier(200, max_depth = 12, random_state = 42, n_jobs = -1, class_weight=\"balanced\"),\n classes = rf_count.classes_,\n per_class = True,\n cmap = \"Set1\")\n\nfig, ax = plt.subplots(figsize=(10, 6))\nax.set_facecolor('#eafff5')\n\nviz.fit(X_train_count,y_train)\nviz.score(X_test_count, y_test)\nviz.show();\n\n# + id=\"1mwPRnTDv9Z9\"\nRF_Count_AP = viz.score_\nRF_Count_AP\n\n# + [markdown] id=\"-u_NPMGhv9Z-\"\n# ### TF-IDF\n\n# + id=\"CwdstzGIv9Z-\"\nrf_tf_idf = RandomForestClassifier(n_estimators = 200, max_depth = 11, class_weight = \"balanced\", random_state = 101, n_jobs = -1)\nrf_tf_idf.fit(X_train_tf_idf, y_train)\n\n# + id=\"OyUMwwNZv9Z-\"\nprint(\"RF_TFIDF Model\")\nprint(\"--------------\")\neval(rf_tf_idf, X_train_tf_idf, X_test_tf_idf)\n\n# + id=\"BDK0ZKhov9Z-\"\nviz = PrecisionRecallCurve(\n RandomForestClassifier(n_estimators = 200, max_depth = 11, class_weight = \"balanced\", random_state = 101, n_jobs = -1),\n classes = rf_tf_idf.classes_,\n per_class = True,\n cmap = \"Set1\")\n\nfig, ax = plt.subplots(figsize=(10, 6))\nax.set_facecolor('#eafff5')\n\nviz.fit(X_train_tf_idf,y_train)\nviz.score(X_test_tf_idf, y_test)\nviz.show();\n\n# + id=\"CVDWts5Yv9Z_\"\ny_pred = rf_tf_idf.predict(X_test_tf_idf)\nrf_tf_idf_rec = recall_score(y_test, y_pred, pos_label = 0, average = None)\nrf_tf_idf_f1 = f1_score(y_test, y_pred, pos_label = 0, average = None)\nrf_tf_idf_AP = viz.score_\n\n# + id=\"GsKba-3av9Z_\"\nprint(\"viz.score_ : \", viz.score_)\nprint(\"RF_TFIDF_rec : \", rf_tf_idf_rec)\nprint(\"RF_TFIDF_f1 : \", rf_tf_idf_f1)\nprint(\"RF_TFIDF_AP : \", rf_tf_idf_AP)\n\n# + [markdown] id=\"LWIJpkCRv9Z_\"\n# ### TF-IDF with Cross Validation\n\n# + id=\"rQw3duU6v9Z_\"\ncustom_scorer = {'accuracy': make_scorer(accuracy_score),\n 'precision-0': make_scorer(precision_score, pos_label = 0),\n 'recall-0': make_scorer(recall_score, pos_label = 0),\n 'f1-0': make_scorer(f1_score, pos_label = 0),\n 'precision-1': make_scorer(precision_score, pos_label = 1),\n 'recall-1': make_scorer(recall_score, pos_label = 1),\n 'f1-1': make_scorer(f1_score, pos_label = 1)\n }\n\nfor i, j in custom_scorer.items():\n model = RandomForestClassifier(n_estimators = 200, max_depth = 11, class_weight = \"balanced\", random_state = 101, n_jobs = -1)\n scores = cross_val_score(model, X_train_tf_idf, y_train, cv = 10, scoring = j).mean()\n if i == \"recall-1\":\n rf_tfidf_rec = scores\n elif i == \"f1-1\":\n rf_tfidf_f1 = scores\n print(f\" {i:20} score for RF_TFIDF : {scores}\\n\")\n\n# + id=\"W868UilYv9Z_\"\nviz = PrecisionRecallCurve(\n RandomForestClassifier(n_estimators = 200, max_depth = 11, class_weight = \"balanced\", random_state = 101, n_jobs = -1),\n classes = rf_tf_idf.classes_,\n per_class = True,\n cmap = \"Set1\")\n\nfig, ax = plt.subplots(figsize = (10, 6))\nax.set_facecolor('#eafff5')\n\nviz.fit(X_train_tf_idf,y_train)\nviz.score(X_test_tf_idf, y_test)\nviz.show();\n\n# + id=\"timSEUwRv9aA\"\nRF_TFIDF_AP = viz.score_\nRF_TFIDF_AP\n\n# + [markdown] id=\"bKbx-Pntv9aA\"\n# ## Ada Boosting\n#\n# ### CountVectorizer\n\n# + id=\"NNgn7ug4v9aA\"\nfrom sklearn.ensemble import AdaBoostClassifier\n\nada_count = AdaBoostClassifier(n_estimators = 500, random_state = 101)\nada_count.fit(X_train_count, y_train)\n\n# + id=\"QAUalcxkv9aA\"\nprint(\"Ada_Count Model\")\nprint(\"---------------\")\neval(ada_count, X_train_count, X_test_count)\n\n# + id=\"G8vDkHGVv9aA\"\nviz = PrecisionRecallCurve(\n AdaBoostClassifier(n_estimators = 500, random_state = 101),\n classes = ada_count.classes_,\n per_class = True,\n cmap = \"Set1\")\n\nfig, ax = plt.subplots(figsize = (10, 6))\nax.set_facecolor('#eafff5')\n\nviz.fit(X_train_count,y_train)\nviz.score(X_test_count, y_test)\nviz.show();\n\n# + id=\"25TqwWJYv9aB\"\ny_pred = ada_count.predict(X_test_count)\nada_count_rec = recall_score(y_test, y_pred, pos_label = 0, average = None)\nada_count_f1 = f1_score(y_test, y_pred, pos_label = 0, average = None)\nada_count_AP = viz.score_\n\n# + id=\"ZwFwYegKv9aB\"\nprint(\"viz.score_ : \", viz.score_)\nprint(\"Ada_Count_rec : \", ada_count_rec)\nprint(\"Ada_Count_f1 : \", ada_count_f1)\nprint(\"Ada_Count_AP : \", ada_count_AP)\n\n# + [markdown] id=\"eJNmYjZlv9aB\"\n# ### CountVectorizer with Cross Validation\n\n# + id=\"FUtoxK6fv9aB\"\ncustom_scorer = {'accuracy': make_scorer(accuracy_score),\n 'precision-0': make_scorer(precision_score, pos_label = 0),\n 'recall-0': make_scorer(recall_score, pos_label = 0),\n 'f1-0': make_scorer(f1_score, pos_label = 0),\n 'precision-1': make_scorer(precision_score, pos_label = 1),\n 'recall-1': make_scorer(recall_score, pos_label = 1),\n 'f1-1': make_scorer(f1_score, pos_label = 1)\n }\n\nfor i, j in custom_scorer.items():\n model = AdaBoostClassifier(n_estimators = 500, random_state = 101)\n scores = cross_val_score(model, X_train_count, y_train, cv = 10, scoring = j).mean()\n if i == \"recall-1\":\n ada_count_rec = scores\n elif i == \"f1-1\":\n ada_count_f1 = scores\n print(f\" {i:20} score for Ada_Count : {scores}\\n\")\n\n# + id=\"vljQ8rjJv9aC\"\nviz = PrecisionRecallCurve(\n AdaBoostClassifier(n_estimators = 500, random_state = 101),\n classes = ada_count.classes_,\n per_class = True,\n cmap = \"Set1\")\n\nfig, ax = plt.subplots(figsize = (10, 6))\nax.set_facecolor('#eafff5')\n\nviz.fit(X_train_count,y_train)\nviz.score(X_test_count, y_test)\nviz.show();\n\n# + id=\"NjMhEEJJv9aC\"\nAda_Count_AP = viz.score_\nAda_Count_AP\n\n# + [markdown] id=\"0A3O62Rwv9aC\"\n# ### TF-IDF\n\n# + id=\"z6IEb_uav9aC\"\nada_tf_idf = AdaBoostClassifier(n_estimators = 500, random_state = 101)\nada_tf_idf.fit(X_train_tf_idf, y_train)\n\n# + id=\"cFfmsZt0v9aC\"\nprint(\"Ada_TFIDF Model\")\nprint(\"---------------\")\neval(ada_tf_idf, X_train_tf_idf, X_test_tf_idf)\n\n# + id=\"nqCHyi_bv9aC\"\nviz = PrecisionRecallCurve(\n AdaBoostClassifier(n_estimators = 500, random_state = 101),\n classes = ada_tf_idf.classes_,\n per_class = True,\n cmap = \"Set1\")\n\nfig, ax = plt.subplots(figsize = (10, 6))\nax.set_facecolor('#eafff5')\n\nviz.fit(X_train_tf_idf,y_train)\nviz.score(X_test_tf_idf, y_test)\nviz.show();\n\n# + id=\"eblMPCFYv9aD\"\ny_pred = ada_tf_idf.predict(X_test_tf_idf)\nada_tf_idf_rec = recall_score(y_test, y_pred, pos_label = 0, average = None)\nada_tf_idf_f1 = f1_score(y_test, y_pred, pos_label = 0, average = None)\nada_tf_idf_AP = viz.score_\n\n# + id=\"mIOiJ54mv9aD\"\nprint(\"viz.score_ : \", viz.score_)\nprint(\"Ada_TFIDF_rec : \", ada_tf_idf_rec)\nprint(\"Ada_TFIDF_f1 : \", ada_tf_idf_f1)\nprint(\"Ada_TFIDF_AP : \", ada_tf_idf_AP)\n\n# + [markdown] id=\"NTIUb_80v9aD\"\n# ### TF-IDF with Cross Validation\n\n# + id=\"VdaHf49_v9aD\"\ncustom_scorer = {'accuracy': make_scorer(accuracy_score),\n 'precision-0': make_scorer(precision_score, pos_label = 0),\n 'recall-0': make_scorer(recall_score, pos_label = 0),\n 'f1-0': make_scorer(f1_score, pos_label = 0),\n 'precision-1': make_scorer(precision_score, pos_label = 1),\n 'recall-1': make_scorer(recall_score, pos_label = 1),\n 'f1-1': make_scorer(f1_score, pos_label = 1)\n }\n\nfor i, j in custom_scorer.items():\n model = AdaBoostClassifier(n_estimators = 500, random_state = 101)\n scores = cross_val_score(model, X_train_tf_idf, y_train, cv = 10, scoring = j).mean()\n if i == \"recall-1\":\n ada_tfidf_rec = scores\n elif i == \"f1-1\":\n ada_tfidf_f1 = scores\n print(f\" {i:20} score for Ada_TFIDF : {scores}\\n\")\n\n# + id=\"R3ZQ8txvv9aD\"\nviz = PrecisionRecallCurve(\n AdaBoostClassifier(n_estimators = 500, random_state = 101),\n classes = ada_tf_idf.classes_,\n per_class = True,\n cmap = \"Set1\")\n\nfig, ax = plt.subplots(figsize = (10, 6))\nax.set_facecolor('#eafff5')\n\nviz.fit(X_train_tf_idf,y_train)\nviz.score(X_test_tf_idf, y_test)\nviz.show();\n\n# + id=\"rKBsgwXyv9aE\"\nAda_TFIDF_AP = viz.score_\nAda_TFIDF_AP\n\n# + [markdown] id=\"T-NYlLNRv9aE\"\n# ## Modelling with Deep Learning\n\n# + id=\"oIrhPIXWv9aE\"\n# df_dl = pd.read_csv(\"Womens Clothing E-Commerce Reviews.csv\")\n# df_dl.head()\n\n# + id=\"WGdwUHfOv9aE\"\ndf_dl = pd.read_csv('../input/womens-ecommerce-clothing-reviews/Womens Clothing E-Commerce Reviews.csv')\n\n# + id=\"qacpxAyDv9aE\"\ndf_dl = df_dl[[\"Review Text\",\"Recommended IND\"]]\ndf_dl.head()\n\n# + id=\"Ff9Uib1Ev9aF\"\ndf_dl.rename(columns = {'Review Text':'text', 'Recommended IND':'recommend'}, inplace = True)\n\n# + id=\"hRjdkL4zv9aF\"\ndf_dl.shape\n\n# + id=\"QIlXN8aCv9aF\"\ndf_dl.isnull().sum()\n\n# + id=\"9l_-vX-2v9aF\"\ndf_dl = df_dl.dropna(subset = ['text'], axis = 0)\ndf_dl = df_dl.reset_index(drop = True)\n\n# + id=\"F3eWL7w9v9aF\"\ndf_dl.isnull().sum()\n\n# + id=\"j0kb0xrZv9aF\"\ndf_dl.shape\n\n# + id=\"gOnkg5Qlv9aG\"\ndf_dl.head()\n\n# + id=\"BrogoUOtv9aG\"\nfrom tensorflow.keras.models import Sequential\nfrom tensorflow.keras.layers import Dense, GRU, Embedding, Dropout\nfrom tensorflow.keras.optimizers import Adam\nfrom tensorflow.keras.preprocessing.text import Tokenizer\nfrom tensorflow.keras.preprocessing.sequence import pad_sequences\nfrom tensorflow.keras.callbacks import EarlyStopping\n\n# + [markdown] id=\"P-h40MODv9aG\"\n# ### Tokenization\n\n# + id=\"TtG8dlwfv9aG\"\nX = df_dl['text'].values\ny = df_dl['recommend'].map({0:1, 1:0}).values\n\n# + id=\"cHCUfqegv9aG\"\nnum_words = 10000\ntokenizer = Tokenizer(num_words = num_words)\n\n# + id=\"4JUOHqxqv9aH\"\ntokenizer.fit_on_texts(X)\n\n# + [markdown] id=\"5QSlxVwav9aH\"\n# ### Creating word index\n\n# + id=\"XVSp0lFVv9aH\"\ntokenizer.word_index\n\n# + id=\"U3uwXBIkv9aH\"\nlen(tokenizer.word_index)\n\n# + id=\"YF8D2_QLv9aH\"\nlen(tokenizer.word_index.keys())\n\n# + [markdown] id=\"pzfytvbKv9aI\"\n# ### Converting tokens to numeric\n\n# + id=\"L_cLKuetv9aI\"\nX_num_tokens = tokenizer.texts_to_sequences(X)\n\n# + id=\"gh_dTcjOv9aI\"\nnum_tokens = [len(tokens) for tokens in X_num_tokens]\nnum_tokens = np.array(num_tokens)\n\n# + id=\"6SCmeliTv9aI\"\nX_num_tokens\n\n# + [markdown] id=\"BTxhknvtv9aI\"\n# ### Maximum number of tokens for all documents¶\n\n# + id=\"bZxEGvQ1v9aJ\"\nnum_tokens.max()\n\n# + id=\"PUorr0UFv9aJ\"\nnum_tokens.mean()\n\n# + id=\"mAFnHfQ8v9aJ\"\nnum_tokens.argmax()\n\n# + id=\"VYSrVAXMv9aJ\"\nX[16263]\n\n# + id=\"z0pVOTXuv9aJ\"\nnum_tokens.argmin()\n\n# + id=\"44F1Re72v9aJ\"\nX[820]\n\n# + [markdown] id=\"JYU_m3cQv9aK\"\n# ### Fixing token counts of all documents (pad_sequences)\n\n# + id=\"Y_cb9rnsv9aK\"\nmax_tokens = 103\n\n# + id=\"gQSsZV1Uv9aK\"\nsum(num_tokens < max_tokens) / len(num_tokens)\n\n# + id=\"10M91NmUv9aK\"\nX_pad = pad_sequences(X_num_tokens, maxlen = max_tokens)\n\n# + id=\"vzwQsjrav9aK\"\nX_pad.shape\n\n# + [markdown] id=\"Gw3XwN3Wv9aK\"\n# ### Train Set Split\n\n# + id=\"BH_k7v5Vv9aL\"\nfrom sklearn.model_selection import train_test_split\nfrom keras.layers import Bidirectional\n\n# + id=\"5hDH5KQrv9aL\"\nX_train, X_test, y_train, y_test = train_test_split(X_pad, y, test_size = 0.2, stratify = y, random_state = 101)\n\n# + [markdown] id=\"UVYl2iArv9aL\"\n# ### Modeling\n\n# + id=\"Xuht9xMPv9aL\"\nmodel = Sequential()\n\n# + id=\"XBMVEXVXv9aL\"\nembedding_size = 100\n\n# + id=\"47TxXBxnv9aL\"\nmodel.add(Embedding(input_dim = num_words,\n output_dim = embedding_size,\n input_length = max_tokens,\n name = 'embedding_layer'))\n\n# + id=\"N0KcFDvwv9aM\"\nmodel.add(Bidirectional(GRU(units = 48, return_sequences = True)))\nmodel.add(Dropout(0.5))\nmodel.add(Bidirectional(GRU(units = 24, return_sequences = True)))\nmodel.add(Dropout(0.5))\nmodel.add(Bidirectional(GRU(units = 12)))\nmodel.add(Dense(1, activation = 'sigmoid'))\n\n# + id=\"EI9JwYdzv9aM\"\noptimizer = Adam(lr = 0.004)\n\n# + id=\"rzLwEAnqv9aM\"\nmodel.compile(loss = 'binary_crossentropy',\n optimizer = optimizer,\n metrics = [\"Recall\"])\n\n# + id=\"Bb2_c9kfv9aN\"\nmodel.summary() \n\n# + id=\"-G-ri9-jv9aO\"\nearly_stop = EarlyStopping(monitor = \"val_loss\", mode = \"min\", \n verbose = 1, patience = 5, restore_best_weights = True)\n\n# + id=\"KMaP41stv9aO\"\npd.Series(y_train).value_counts(normalize = True) \n\n# + id=\"9A3MnFUsv9aO\"\nweights = {0:19, 1:81}\n\n# + id=\"42STv2nMv9aO\"\nmodel.fit(X_train, y_train, epochs = 25, batch_size = 256, class_weight = weights,\n validation_data = (X_test, y_test), callbacks = [early_stop])\n\n# + [markdown] id=\"lTEefs6av9aO\"\n# ### Model evaluation\n\n# + id=\"3hVK_fNZv9aO\"\nmodel_loss = pd.DataFrame(model.history.history)\nmodel_loss.head()\n\n# + id=\"uaY8joiDv9aP\"\nmodel_loss.plot()\n\n# + id=\"3bhVfi5Hv9aP\"\nmodel.evaluate(X_train, y_train)\n\n# + id=\"lM02jS1zv9aP\"\nmodel.evaluate(X_test, y_test) \n\n# + id=\"DEonHhMgv9aP\"\nfrom sklearn.metrics import confusion_matrix, classification_report, accuracy_score, f1_score, roc_auc_score\n\ny_train_pred = (model.predict(X_train) >= 0.5).astype(\"int32\")\n\nprint(confusion_matrix(y_train, y_train_pred))\nprint(\"---------------------------------------------\")\nprint(classification_report(y_train, y_train_pred))\n\n# + id=\"6bxjUppvv9aP\"\ny_pred = (model.predict(X_test) >= 0.5).astype(\"int32\")\n\nprint(confusion_matrix(y_test, y_pred))\nprint(\"--------------------------------------\")\nprint(classification_report(y_test, y_pred))\n\n# + id=\"LcnEY6DRv9aP\"\nfrom sklearn.metrics import precision_recall_curve, average_precision_score\n\n# + id=\"fnsUTH1jv9aQ\"\ny_pred_proba = model.predict(X_test)\nprecision, recall, thresholds = precision_recall_curve(y_test, y_pred_proba)\nplt.plot([0,1],[0,1],'k--')\nplt.plot(precision, recall)\nplt.xlabel('precision')\nplt.ylabel('recall')\nplt.title('Precision Recall Curve')\nplt.show()\n\n# + id=\"VazBLEVMv9aQ\"\nDL_AP = average_precision_score(y_test, y_pred_proba)\nDL_f1 = f1_score(y_test, y_pred)\nDL_rec = recall_score(y_test, y_pred)\n\n# + id=\"roW942Bpv9aQ\"\nprint(\"DL_AP : \", DL_AP)\nprint(\"DL_f1 : \", DL_f1)\nprint(\"DL_rec : \", DL_rec)\n\n# + [markdown] id=\"Hdc5JYOXCV4A\"\n# ### Compare Models F1 Scores, Recall Scores and Average Precision Score\n\n# + id=\"PEqjib60v9aQ\"\ncompare = pd.DataFrame({\"Model\": [\"NaiveBayes(Multi)_Count\", \"NaiveBayes(Berno)_Count\", \"LogReg_Count\", \"SVM_Count\", \n \"Random Forest_Count\", \"AdaBoost_Count\", \"NaiveBayes(Multi)_TFIDF\", \n \"NaiveBayes(Berno)_TFIDF\", \"LogReg_TFIDF\", \"SVM_TFIDF\", \"Random Forest_TFIDF\", \n \"AdaBoost_TFIDF\", \"DL\"],\n \n \"F1_Score\": [nbm_count_f1, nbb_count_f1, log_count_f1, svc_count_f1, rf_count_f1, ada_count_f1,\n nbm_tfidf_f1, nbb_tfidf_f1, log_tfidf_f1, svc_tfidf_f1, rf_tfidf_f1, ada_tfidf_f1, DL_f1],\n \n \"Recall_Score\": [nbm_count_rec, nbb_count_rec, log_count_rec, svc_count_rec, rf_count_rec, \n ada_count_rec, nbm_tfidf_rec, nbb_tfidf_rec, log_tfidf_rec, svc_tfidf_rec, \n rf_tfidf_rec, ada_tfidf_rec, DL_rec],\n \n \"Average_Precision_Score\": [NBMulti_Count_AP, NBBerno_Count_AP, LogReg_Count_AP, SVC_Count_AP, \n RF_Count_AP, Ada_Count_AP, NBMulti_TFIDF_AP, NBBerno_TFIDF_AP,\n LogReg_TFIDF_AP, SVC_TFIDF_AP, RF_TFIDF_AP, Ada_TFIDF_AP, DL_AP]})\n \n \ncompare = compare.sort_values(by=\"Recall_Score\", ascending=True)\nfig = px.bar(compare, x = \"Recall_Score\", y = \"Model\", title = \"Recall_Score\")\nfig.show()\n\ncompare = compare.sort_values(by=\"F1_Score\", ascending=True)\nfig = px.bar(compare, x = \"F1_Score\", y = \"Model\", title = \"F1_Score\")\nfig.show()\n\ncompare = compare.sort_values(by=\"Average_Precision_Score\", ascending=True)\nfig = px.bar(compare, x = \"Average_Precision_Score\", y = \"Model\", title = \"Average_Precision_Score\")\nfig.show()\n\n# + [markdown] id=\"S9QZGHkZv9aR\"\n# ## Prediction\n\n# + id=\"lTN7Ozxvv9aR\"\nreview1 = \"I ordered this dress in 0p since i am 5ft. it fits great its not too short just above the knee.\"\nreview2 = \"Love the top design, not so much the fabric. i should've read the top was 100% polyester. returning and finding a better quality top.\"\nreview3 = \"Love this henely, the lace-up is cute. the shirt is comfy. so versatile. beach coverup or over jeans\"\nreview4 = \"Loved this dress online, but my small pettie stature made me looks like an orphan from the prairie.\"\nreview5 = \"I tried it in the store but was not the true size so i ordered online. i wore it only once so far and got lots of compliments. very unique design.\"\nreview6 = \"Nice shirt seems well made. good just not a great fit for me.\"\nreview7 = \"I tried on this dress in store and was amazed by the quality and simple structure of the dress. bought it with no hesitation.\"\nreview8 = \"This top has a bit of a retro flare but so adorable on. looks really cute with a pair of faded boot cut jeans.\"\nreview9 = \"Great quality top. i do wish it fit me...i'm really long waisted and this top was too short. had to return\"\nreview10 = \"A serious joke. i struggled with the buttons for a good 10 minutes and gave up after the 3rd button. i'm not sure what they were thinking.\"\n\nreviews = [review1, review2, review3, review4, review5, review6, review7, review8, review9, review10]\n\n# + id=\"HbN-QQ8tv9aR\"\ntokens = tokenizer.texts_to_sequences(reviews) \n\n# + id=\"xr3itPICv9aR\"\ntokens_pad = pad_sequences(tokens, maxlen = max_tokens)\ntokens_pad.shape\n\n# + id=\"jRXtORNXv9aS\"\nmod_pred = model.predict(tokens_pad)\n\n# + id=\"0z72cWXDv9aS\"\nmod_pred\n\n# + id=\"mnmBjqJyv9aS\"\ndf_pred = pd.DataFrame(mod_pred, index = reviews)\ndf_pred.rename(columns = {0:'Pred_Proba'}, inplace = True)\n\n# + id=\"Y8ef9DzXv9aS\"\ndf_pred[\"Predicted_Feedbaack\"] = df_pred[\"Pred_Proba\"].apply(lambda x: \"Not Recommended\" if x >= 0.5 else \"Recommended\")\n\n# + id=\"3D-Zfzdkv9aS\"\ndf_pred\n\n# + [markdown] id=\"mfUL4ZaXCV4A\"\n# ### Conclusion\n\n# + [markdown] id=\"n2Ty7VKTv9aT\"\n# - In this project we have used sentiment analysis to determine whether the product is recommended or not. We have built models with five different machine learning algorithms and also with deep learning algorithm and compare their performance. Thus, we have determined the algorithm that makes the most accurate emotion estimation by using the information obtained from the * Review Text * variable.\n#\n# - When the scores are examined in the Compare section, it is seen that the scores are generally close to each other, but the Logistic Regression model stands out. \n#\n# - Thank you in advance for your constructive and instructive comments.\n\n# + id=\"KwLFshu7v9aT\"\n\n","repo_name":"aysedata/NLP","sub_path":"NLP__SENTIMENT_ANALYSIS_PROJECT_aysedata.ipynb","file_name":"NLP__SENTIMENT_ANALYSIS_PROJECT_aysedata.ipynb","file_ext":"py","file_size_in_byte":73732,"program_lang":"python","lang":"en","doc_type":"code","stars":0,"dataset":"github-jupyter-script","pt":"44"}
+{"seq_id":"31809237673","text":"# + id=\"PpOhaGbfbmx8\"\n# 문자열안에 문자열\n# 문자열 str1, str2가 매개변수로 주어집니다.\n# str1 안에 str2가 있다면 1을 없다면 2를 return하도록 solution 함수를 완성해주세요.\n\ndef solution(str1, str2):\n if str2 in str1:\n return 1\n else:\n return 2\n\n\n# + id=\"8c_WE2tkjH_U\"\n# 소인수분해\n# 소인수분해란 어떤 수를 소수들의 곱으로 표현하는 것입니다. 예를 들어 12를 소인수 분해하면 2 * 2 * 3 으로 나타낼 수 있습니다.\n# 따라서 12의 소인수는 2와 3입니다. 자연수 n이 매개변수로 주어질 때 n의 소인수를 오름차순으로 담은 배열을 return하도록 solution 함수를 완성해주세요.\n\ndef solution(n):\n answer = []\n i = 2\n while i <= n:\n if n % i == 0:\n answer.append(i)\n n = n // i\n else:\n i += 1\n return sorted(list(set(answer))) # set을 사용하여 중복된 숫자를 하나의 숫자로 만들고 오름차순으로\n","repo_name":"yoonji3540/programmers-level0","sub_path":"프로그래머스0단계_0529/프로그래머스0단계_0529.ipynb","file_name":"프로그래머스0단계_0529.ipynb","file_ext":"py","file_size_in_byte":1023,"program_lang":"python","lang":"ko","doc_type":"code","stars":0,"dataset":"github-jupyter-script","pt":"44"}
+{"seq_id":"3103015655","text":"# +\ncd = []\n\n\ndef show_cd():\n if len(cd)<=0:\n print(\"masukan data baru\")\n else:\n for indeks in range(len(cd)):\n print(\"\", (indeks, cd[indeks]))\n pilih_menu() \n\ndef insert_cd():\n cd_baru = input(\"nama cd baru: \")\n cd.append(cd_baru)\n\n pilih_menu()\n\ndef edit_cd():\n if(indeks > len(cd)):\n print(\"id salah\")\n else:\n for i in range(len(cd)):\n nama_baru = input(\"edit judul cd: \", cd(i))\n cd = nama_baru\n \n show_cd()\n \n \ndef delete_cd():\n show_cd()\n indeks = input(\"masukan id cd: \")\n if(indeks > len(cd)):\n print(\"id salah\")\n else:\n cd.remove(cd[indeks])\n \n menu = input(\"PILIH MENU> \")\n pilih_menu()\n \ndef show_menu():\n print(\"FLOODRIVER CD STORE\")\n print(\"--------------------\")\n print(\"[1] Lihat CD\")\n print(\"[2] Masukan CD\")\n print(\"[3] Edit CD\")\n print(\"[4] Hapus CD\")\n \ndef pilih_menu():\n menu = input(\"PILIH MENU> \")\n if(menu==\"1\"):\n print(\"[1] Lihat CD\")\n show_cd()\n elif (menu==\"2\"):\n print(\"[2] Masukan CD\")\n insert_cd()\n elif(menu==\"3\"):\n print(\"[3] Edit CD\")\n edit_cd()\n elif(menu==\"4\"):\n print(\"[4] Hapus CD\")\n delete_cd()\n exit()\n else:\n print(\"Salah pilih bos\")\n \n pilih_menu()\n\nif(True):\n show_menu()\n pilih_menu()\n \n \n# -\n\n\n","repo_name":"ridwanwimas1995/Latihan-CRUD","sub_path":"Jual Beli CD Python .ipynb","file_name":"Jual Beli CD Python .ipynb","file_ext":"py","file_size_in_byte":1406,"program_lang":"python","lang":"en","doc_type":"code","stars":0,"dataset":"github-jupyter-script","pt":"44"}
+{"seq_id":"14777884479","text":"# + [markdown] id=\"view-in-github\" colab_type=\"text\"\n# \n\n# + [markdown] id=\"1iO_ZrIG-Xr8\"\n# \n\n# + id=\"WJ4OBiis-Xr9\"\n## Importing Basic Package\nimport numpy as np\nimport pandas as pd\npd.set_option(\"display.max_columns\", 100)\n\n## Importing Visualization\nimport matplotlib.pyplot as plt\n# %matplotlib inline\nimport seaborn as sns\n\n## Importing SKLearn Libraries\nfrom sklearn.model_selection import train_test_split\nfrom sklearn import metrics\nfrom sklearn.preprocessing import StandardScaler\nfrom sklearn.preprocessing import PowerTransformer\n\nfrom sklearn.model_selection import KFold\nfrom sklearn.model_selection import GridSearchCV\nfrom sklearn.linear_model import LogisticRegression\nfrom sklearn.tree import DecisionTreeClassifier\nfrom xgboost import XGBClassifier\n\nfrom sklearn.metrics import roc_curve, roc_auc_score\nfrom sklearn.metrics import f1_score, classification_report\n\nimport warnings\nwarnings.filterwarnings(\"ignore\")\n\n# + colab={\"base_uri\": \"https://localhost:8080/\"} id=\"uyaO9vIxkZ21\" outputId=\"b7d7d3d0-af42-4f0f-c0c2-b82615fc7640\"\n## Load the data from the Google Drive\nfrom google.colab import drive\ndrive.mount('/content/drive')\ndf = pd.read_csv('/content/drive/MyDrive/Colab Notebooks/notebook_data/creditcard.csv')\n\n# + [markdown] id=\"0kTzyqDo-Xr-\"\n# \n# 2. Identify if \"balancing\" the data improve the classification performance \n# 3. AUC-ROC score to be used to measure how well the model is performing \n# 4. 0.172% Fruad, 99.828% Non - Fraud Trasnactions \n# 5. Split the data in 2 parts Train and Test \n# 6. Train/Test Dataset Split (393 / 99) \n#\n\n# + id=\"K6ukUgL8-XsA\" colab={\"base_uri\": \"https://localhost:8080/\", \"height\": 487} outputId=\"89a7d4cb-84bb-4def-e6d4-e008ed6635c9\"\n## Bar Plot for Percentage fraudulent vs non-fraudulent\n## Used Log Scale due to Highly Imbalanced data\nfig = plt.figure(figsize = (5, 5))\n\n\n# creating the bar plot\nbars = [\"Normal Tx\", \"Fraud Tx\"]\nvalues = [normal_share, fraud_share]\nplt.bar(bars, values, color = \"#2D5B6B\", width = 0.4, edgecolor=\"cyan\")\n\nplt.xlabel(\"Percentage of Normal vs Fraud Transactions (Log Scale)\")\nplt.yscale('log')\nplt.ylabel(\"Percentage\")\nplt.title(\"Transactions\")\nplt.show()\n\n\n# + colab={\"base_uri\": \"https://localhost:8080/\", \"height\": 487} id=\"GFgRNgkspVBV\" outputId=\"5d4f2448-7c1c-4306-d5ef-a577abb6a4df\"\n## Bar Plot for count of fraudulent vs non-fraudulent\n## Used Log Scale due to Highly Imbalanced data\n\nfig = plt.figure(figsize = (5, 5))\n\n# creating the bar plot\nbars = [\"Normal Tx\", \"Fraud Tx\"]\nvalues = [normal_share_num, fraud_share_num]\nplt.bar(bars, values, color = \"#7B7D2A\", width = 0.4, edgecolor=\"cyan\")\n\nplt.xlabel(\"Number of Normal vs Fraud Transactions (Log Scale)\")\nplt.ylabel(\"Number of Transactions\")\nplt.yscale('log')\nplt.title(\"Tranactions\")\nplt.show()\n\n\n# + [markdown] id=\"i8JD9L8o32iD\"\n#\n\n# + colab={\"base_uri\": \"https://localhost:8080/\", \"height\": 472} id=\"CrTDJd_I4yrL\" outputId=\"2e9db949-ff64-40e1-c05a-02dea469daad\"\n## Scatter plot to observe the distribution of Class vs Time\n\nsns.scatterplot(data = df, x=\"Time\", y=\"Class\", hue=\"Time\")\nplt.title(\"Time vs Class scatter plot\")\nplt.show()\n\n# + colab={\"base_uri\": \"https://localhost:8080/\", \"height\": 472} id=\"DNzkk6nN4ygX\" outputId=\"87cec4b6-46f7-469e-ea87-fe3e21305137\"\n## Scatter plot to observe the distribution of Class vs Amount\n\nsns.scatterplot(data = df, x=\"Amount\", y=\"Class\", hue=\"Amount\")\nplt.title(\"Time vs Amount scatter plot\")\nplt.show()\n\n# + [markdown] id=\"pV-DDZw938bm\"\n# No specific insight can be drwan from the distribution of the fraudulent transaction \n# Drop Time Feature as it doesn't add any value \n\n# + id=\"vRGWxbAR6_7e\"\ndf = df.drop(\"Time\", axis = 1)\n\n# + [markdown] id=\"O6DOztNd7ROG\"\n# Plotting the distributions of all the featuers\n\n# + colab={\"base_uri\": \"https://localhost:8080/\", \"height\": 1000} id=\"XNFkyJ6c7ehN\" outputId=\"84d82f87-0b1c-4aa2-e7a9-3b12b023f31d\"\n# Plotting all the variable in displot to visualise the distribution\nvar = list(df.columns.values)\n# dropping Class columns from the list\nvar.remove(\"Class\")\n\ni = 0\nt0 = df.loc[df['Class'] == 0]\nt1 = df.loc[df['Class'] == 1]\n\nplt.figure()\nfig, ax = plt.subplots(8,4,figsize=(10,20))\n\nfor feature in var:\n i += 1\n plt.subplot(8,4,i)\n sns.kdeplot(t0[feature], bw=0.5,label=\"0\")\n sns.kdeplot(t1[feature], bw=0.5,label=\"1\")\n plt.xlabel(feature, fontsize=12)\n locs, labels = plt.xticks()\n plt.tick_params(axis='both', which='major', labelsize=12)\nplt.show();\n\n# + [markdown] id=\"PfInMCPe-XsB\"\n# Splitting the data into train & test data\n\n# + id=\"lZnhgQkC-XsB\" colab={\"base_uri\": \"https://localhost:8080/\"} outputId=\"7832abf0-0d60-415e-b44f-76b5bee70f90\"\ny= df[\"Class\"]\nX = df.drop(\"Class\", axis = 1)\n## Spltting the into 80:20 train test size\nX_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.2, random_state = 42)\n## Checking the split of the class lable\nprint(\"Total Class Lables\", np.sum(y))\nprint(\"Train Class Lables\", np.sum(y_train))\nprint(\"Test Class Lables\", np.sum(y_test))\n\n# + [markdown] id=\"nGzOM3SdJ921\"\n# The dataset is PCA transformed \n# Following models may be evaluated later if good perfomance is not high (satisfactory) as they are computationally expensive 1. SVM 2. RandomForest\n\n# + id=\"NjWTNPbL-XsD\" colab={\"base_uri\": \"https://localhost:8080/\"} outputId=\"f1df5917-876b-4143-d9fb-c7076ba6115f\"\n## Logistic Regression parameters for K-fold cross vaidation\nparams = {\"C\": [0.01, 0.1, 1, 10, 100, 1000]}\nfolds = KFold(n_splits=5, shuffle=True, random_state=4)\n\n\n## Cross validation\nmodel_cv = GridSearchCV(estimator = LogisticRegression(),\n param_grid = params,\n scoring= 'roc_auc',\n cv = folds,\n n_jobs=-1,\n verbose = 1,\n return_train_score=True)\n## hyperparameter tuning\nmodel_cv.fit(X_train, y_train)\n\n## Output of the evaluation result\nprint('Best ROC AUC score: ', model_cv.best_score_)\n\n## Output of the best hyperparametes\nprint('Best hyperparameters: ', model_cv.best_params_)\n\n\n# + id=\"fV8emcSDC_ty\" colab={\"base_uri\": \"https://localhost:8080/\", \"height\": 371} outputId=\"8b13875e-653d-46ac-d0d1-37b3544fd392\"\n## Cross Validation results\ncv_results = pd.DataFrame(model_cv.cv_results_)\ncv_results\n\n# + id=\"rwsxTzekDJ3M\" colab={\"base_uri\": \"https://localhost:8080/\", \"height\": 546} outputId=\"186b3a45-d61b-4c92-ab73-33a0202b382b\"\n## C Vs. Train and Validation scores\nplt.figure(figsize=(8, 6))\nplt.plot(cv_results['param_C'], cv_results['mean_test_score'])\nplt.plot(cv_results['param_C'], cv_results['mean_train_score'])\nplt.xlabel('C')\nplt.ylabel('sensitivity')\nplt.legend(['test result', 'train result'], loc='upper left')\nplt.xscale('log')\n\n# + id=\"2_erEYeuD4bb\" colab={\"base_uri\": \"https://localhost:8080/\", \"height\": 75} outputId=\"8d8349c1-c035-4d17-c5b8-db7a7ab39196\"\n\n## Logistic Regression with optimal C\n\n## Instantiating the model with best C\nlog_reg_imb_model = LogisticRegression(C=0.01)\n\n## Fitting the model on train dataset\nlog_reg_imb_model.fit(X_train, y_train)\n\n\n# + id=\"0jDvg8nQD4Yl\"\n## Prediction and model evalution on the train set\n# Creating function to display ROC-AUC score, f1 score and classification report\ndef display_scores(y_test, y_pred):\n '''\n Display ROC-AUC score, f1 score and classification report of a model.\n '''\n print(f\"F1 Score: {round(f1_score(y_test, y_pred)*100,2)}%\")\n print(f\"Classification Report: \\n {classification_report(y_test, y_pred)}\")\n\n\n# + id=\"r0g7ik97D4V8\"\ny_train_pred = log_reg_imb_model.predict(X_train)\n\n# + id=\"uaBbXtaJD4Sn\" colab={\"base_uri\": \"https://localhost:8080/\"} outputId=\"fb902c05-a43a-4767-d9e7-126e09340322\"\ndisplay_scores(y_train, y_train_pred)\n\n\n# + id=\"EiFUybELD4Ps\"\n# ROC Curve function\ndef draw_roc( actual, probs ):\n fpr, tpr, thresholds = metrics.roc_curve( actual, probs,\n drop_intermediate = False )\n auc_score = metrics.roc_auc_score( actual, probs )\n plt.figure(figsize=(5, 5))\n plt.plot( fpr, tpr, label='ROC curve (area = %0.2f)' % auc_score )\n plt.plot([0, 1], [0, 1], 'k--')\n plt.xlim([0.0, 1.0])\n plt.ylim([0.0, 1.05])\n plt.xlabel('False Positive Rate or [1 - True Negative Rate]')\n plt.ylabel('True Positive Rate')\n plt.title('Receiver operating characteristic example')\n plt.legend(loc=\"lower right\")\n plt.show()\n\n return None\n\n\n# + id=\"RWxcwt3hD4ND\"\n## Predicted probability\ny_train_pred_proba = log_reg_imb_model.predict_proba(X_train)[:,1]\n\n# + id=\"5WWj3odkD4J9\" colab={\"base_uri\": \"https://localhost:8080/\", \"height\": 487} outputId=\"23e9d53c-13c5-4576-9529-15c549d53ec1\"\n## Plot the ROC curve\ndraw_roc(y_train, y_train_pred_proba)\n\n# + id=\"osJMibPOD4HK\" colab={\"base_uri\": \"https://localhost:8080/\"} outputId=\"06a18d0c-eabf-4e1e-80e1-b24a3361cc1b\"\n##Evaluating the model on the test set\ny_test_pred = log_reg_imb_model.predict(X_test)\ndisplay_scores(y_test, y_test_pred)\n\n# + id=\"7kfQOZRbD4EW\"\n# Predicted probability\ny_test_pred_proba = log_reg_imb_model.predict_proba(X_test)[:,1]\n\n# + id=\"d6nPf6qqD4Bb\" colab={\"base_uri\": \"https://localhost:8080/\", \"height\": 487} outputId=\"e9f6f46a-e9bc-44d7-8a9c-368e78145237\"\n# Plot the ROC curve\ndraw_roc(y_test, y_test_pred_proba)\n\n# + [markdown] id=\"vVNVI2rWE96U\"\n# Model Summary\n#\n# ### Train set\n# - ROC : 98%\n# - F1 Score: 73.76%\n#\n# ### Test set\n# - ROC : 98%\n# - F1 score: 70.3%\n\n# + [markdown] id=\"9XQRZ-r_FJBg\"\n# Decision Tree\n\n# + id=\"lp2PxQVeD3-g\" colab={\"base_uri\": \"https://localhost:8080/\", \"height\": 135} outputId=\"c171c679-1556-4e7e-b557-85b047f87d4c\"\n# Create the parameter grid\nparam_grid = {\n 'max_depth': range(5, 15, 5),\n 'min_samples_leaf': range(50, 150, 50),\n 'min_samples_split': range(50, 150, 50),\n}\n\n\n# Instantiate the grid search model\ndtree = DecisionTreeClassifier()\n\ngrid_search = GridSearchCV(estimator = dtree,\n param_grid = param_grid,\n scoring= 'roc_auc',\n cv = 3,\n n_jobs=-1,\n verbose = 1)\n\n# Fit the grid search to the data\ngrid_search.fit(X_train,y_train)\n\n# + id=\"QWdzXsB6D37s\" colab={\"base_uri\": \"https://localhost:8080/\", \"height\": 642} outputId=\"09ddb1ec-f557-4fb6-c8fa-bd614242c12a\"\n# cv results\ncv_results = pd.DataFrame(grid_search.cv_results_)\ncv_results\n\n# + id=\"DXDm0m9GD34x\" colab={\"base_uri\": \"https://localhost:8080/\"} outputId=\"816dbaea-8f7e-4ded-9cb3-b71c693c8f26\"\n# Printing the optimal score and hyperparameters\nprint(\"Best roc auc score : \", grid_search.best_score_)\nprint(grid_search.best_estimator_)\n\n# + id=\"zZ6901iRD316\" colab={\"base_uri\": \"https://localhost:8080/\", \"height\": 92} outputId=\"d2bf3e8f-dfe0-4589-dafd-3b082cfeede3\"\n## Decision Tree with optimal hyperparameters\n# Model with optimal hyperparameters\ndt_imb_model = DecisionTreeClassifier(criterion = \"gini\",\n random_state = 100,\n max_depth=10,\n min_samples_leaf=100,\n min_samples_split=100)\n\ndt_imb_model.fit(X_train, y_train)\n\n# + id=\"AcTWfsiUF5lJ\" colab={\"base_uri\": \"https://localhost:8080/\"} outputId=\"da20a5d7-7752-4435-c7b4-a0f8897d3248\"\n## Prediction on the train set\ny_train_pred = dt_imb_model.predict(X_train)\ndisplay_scores(y_train, y_train_pred)\n\n# + id=\"hR9GpTXcF5iT\" colab={\"base_uri\": \"https://localhost:8080/\", \"height\": 487} outputId=\"3f47ff8f-f0ac-46e7-bc05-1342da4b2c71\"\n# Predicted probability\ny_train_pred_proba = dt_imb_model.predict_proba(X_train)[:,1]\n\n# Plot the ROC curve\ndraw_roc(y_train, y_train_pred_proba)\n\n# + id=\"QEyvXynKGL7R\" colab={\"base_uri\": \"https://localhost:8080/\"} outputId=\"67cfde55-43b9-4a98-8d43-1b7e2e0511e4\"\n## Evaluating the model on the test set\ny_test_pred = dt_imb_model.predict(X_test)\ndisplay_scores(y_test, y_test_pred)\n\n# + id=\"y2M1jvVNGL36\" colab={\"base_uri\": \"https://localhost:8080/\", \"height\": 487} outputId=\"c21faf72-0ce9-498a-ebd8-cb531f52d678\"\n# Predicted probability\ny_test_pred_proba = dt_imb_model.predict_proba(X_test)[:,1]\n\n# Plot the ROC curve\ndraw_roc(y_test, y_test_pred_proba)\n\n# + [markdown] id=\"RDorE5L3GXIS\"\n# Model Summary\n#\n# Train set\n# - ROC Score: 99%\n# - F1 score : 72.33% \n#\n# Test set\n# - ROC Score: 96%\n# - F1 score : 71.88%\n#\n# \n# - Executed Random Oversampling, SMOTE, and ADASYN technique to balance the dataset\n# - Modeling using logistic regression, random forest and XGBoost algorithms\n#\n# **Conclusion: *XGBoost* model is performing well on the dataset which is balanced with AdaSyn technique.\n# Hence, we conclude that the XGBoost model with Adasyn is the best model.**\n\n# + colab={\"base_uri\": \"https://localhost:8080/\", \"height\": 1000} id=\"aov1Xo0swfV1\" outputId=\"4aaf53cf-eb8a-4a45-f8ef-1074d56d7741\"\nvar_imp = []\nfor i in xgb_adasyn_model.feature_importances_:\n var_imp.append(i)\nprint('Top var =', var_imp.index(np.sort(xgb_adasyn_model.feature_importances_)[-1])+1)\nprint('2nd Top var =', var_imp.index(np.sort(xgb_adasyn_model.feature_importances_)[-2])+1)\nprint('3rd Top var =', var_imp.index(np.sort(xgb_adasyn_model.feature_importances_)[-3])+1)\n\n# Variable on Index-13 and Index-9 seems to be the top 2 variables\ntop_var_index = var_imp.index(np.sort(xgb_adasyn_model.feature_importances_)[-1])\nsecond_top_var_index = var_imp.index(np.sort(xgb_adasyn_model.feature_importances_)[-2])\n\nX_train_1 = X_train.to_numpy()[np.where(y_train==1.0)]\nX_train_0 = X_train.to_numpy()[np.where(y_train==0.0)]\n\nnp.random.shuffle(X_train_0)\n\nimport matplotlib.pyplot as plt\n# %matplotlib inline\nplt.rcParams['figure.figsize'] = [20, 20]\n\nplt.scatter(X_train_1[:, top_var_index], X_train_1[:, second_top_var_index], label='Actual Class-1 Examples')\nplt.scatter(X_train_0[:X_train_1.shape[0], top_var_index], X_train_0[:X_train_1.shape[0], second_top_var_index],\n label='Actual Class-0 Examples')\nplt.legend()\n\n# + colab={\"base_uri\": \"https://localhost:8080/\"} id=\"jAylklngujHM\" outputId=\"7eac3f83-e3ed-4288-ff6a-591d8331d455\"\nprint('Train auc =', metrics.roc_auc_score(y_train_adasyn, y_train_pred_proba))\nfpr, tpr, thresholds = metrics.roc_curve(y_train_adasyn, y_train_pred_proba )\nthreshold = thresholds[np.argmax(tpr-fpr)]\nprint(threshold)\n\n# + [markdown] id=\"WYeze4yvxkt3\"\n# ***Threshold: 89% at which TPR is the highest and FPR is the lowest and we get 100% ROC score on the train data.***\n\n# + [markdown] id=\"yRKmtyCVjiye\"\n#\n#\n# ---\n# > End of Report\n#\n","repo_name":"taruj/capstone","sub_path":"Taruj_CreditCard_Fraud_Detection_Capstone.ipynb","file_name":"Taruj_CreditCard_Fraud_Detection_Capstone.ipynb","file_ext":"py","file_size_in_byte":60126,"program_lang":"python","lang":"en","doc_type":"code","stars":0,"dataset":"github-jupyter-script","pt":"44"}
+{"seq_id":"38527488465","text":"# ### Stratisfied Sampling based on sentence length\n\nimport pandas as pd\npd.set_option('display.max_colwidth', None)\n\n# +\ndir_path = \"./\"\noutprefix = \"./data-\"\nsample_size = 100\n\ndf = pd.read_csv(f\"{dir_path}/test.csv\")\n\ndf['slen'] = df.sentence.apply(lambda x : len(str(x).split()))\nn = len(df)\nprint(n)\ndf.head()\n# -\n\nsample_1 = df.groupby('slen', group_keys=False).apply(lambda x: x.sample(round(len(x)/n*sample_size)))\nprint(len(sample_1))\nsample_1.head()\n\n# +\ndf_res = df[(~df.slen.isin(sample_1.slen))]\ndf_res = df_res.groupby('slen', group_keys=False).apply(lambda x: x.sample(1))\nn_res = sample_size - len(sample_1)\nprint(n_res)\n\nsample_2 = df_res.sample(n_res)\nsample_2.head()\n\n# +\ndf_infer = pd.concat([sample_1, sample_2])\ndf_infer = df_infer.sample(frac=1)\n\ndf_infer.to_csv(f\"{outprefix}infer.csv\")\nwith open(f\"{outprefix}infer.txt\", \"w+\") as fw:\n for i, row in df_infer.iterrows():\n fw.write(f'{row[\"sentence\"]}\\n')\n","repo_name":"madenindya/my_scripts","sub_path":"statisfied_sampling.ipynb","file_name":"statisfied_sampling.ipynb","file_ext":"py","file_size_in_byte":939,"program_lang":"python","lang":"en","doc_type":"code","stars":0,"dataset":"github-jupyter-script","pt":"44"}
+{"seq_id":"25493305976","text":"# + [markdown] id=\"eoHcpUm_XvJb\"\n# ### импорты\n#\n\n# + colab={\"base_uri\": \"https://localhost:8080/\"} id=\"cRo15Yj8kX0l\" outputId=\"c3ee63fa-f47b-4e16-9e40-0d7ef16f89f0\"\n# !pip install jira\n# !pip install pymorphy2\n\n# + id=\"CI-onWZDWQLa\"\nimport numpy as np\nimport pandas as pd\nimport tensorflow as tf\n\nfrom jira import JIRA\nimport json\nimport ast\nimport pymorphy2\n\nfrom sklearn.preprocessing import LabelEncoder\nfrom sklearn.metrics import confusion_matrix, accuracy_score\n\nimport matplotlib.pyplot as plt\nimport seaborn as sns\nsns.set()\nplt.style.use('ggplot')\nplt.rcParams['figure.figsize'] = (10,10)\n\nFILE_PATH = '/content/drive/MyDrive/Colab Notebooks/Korus/nlp_tasks/find_assignee/data/'\n\n\n# + [markdown] id=\"_YO9Fb5MWpjp\"\n# ### Загрузить модель\n#\n\n# + colab={\"base_uri\": \"https://localhost:8080/\"} id=\"J9rummBNWbn7\" outputId=\"00a09a97-f411-4ac2-cefa-cbd277559050\"\nfile_name = 'epoch_29_val_acc_0.80.hdf5'\nmodel = tf.keras.models.load_model(FILE_PATH+file_name)\nmodel.summary()\n\n# + id=\"vDmTEMbndsd4\"\n\n\n# + [markdown] id=\"IxZpgO6FWs-d\"\n# ### Посмотреть данные, на которых обучалась модель\n\n# + colab={\"base_uri\": \"https://localhost:8080/\", \"height\": 320} id=\"cPciB7yEWyOU\" outputId=\"0ea5fcee-97ef-43fe-f008-32351cb44941\"\n#тренировочные данные\nfile_name = 'find_assignee_010620-260421_for_training.csv'\ntraining_data = pd.read_csv(FILE_PATH+file_name, sep=';', index_col='issue_number')\ntraining_data['full_text'] = training_data['full_text'].apply(ast.literal_eval)\ntraining_data.head(5)\n\n# + colab={\"base_uri\": \"https://localhost:8080/\", \"height\": 320} id=\"1D4pkQmJFHKI\" outputId=\"1580c434-a2fb-4fc7-8822-515fc0e02b16\"\n# # без предобработки\n# file_name = 'find_assignee_010620-260421_validated.csv' \n# training_data = pd.read_csv(FILE_PATH+file_name, sep=';', index_col='issue_number')\n# training_data.head(5)\n\n# + [markdown] id=\"-5GxHQ8pWyWC\"\n# ### получить несколько запросов\n\n# + id=\"rU6kj1gFizA5\"\nimport json\n\n\ndef read_json(file_path):\n with open(file_path, \"r\") as f:\n return json.load(f)\n\n\n#читаю конфиг\ncfg = read_json(FILE_PATH+\"cfg.json\")\njira_options = {'server': cfg['server']}\nbasic_auth=(cfg['username'], cfg['password'])\n\n# Открываю сессию jira\njira = JIRA(options=jira_options, basic_auth=basic_auth)\n\n# + colab={\"base_uri\": \"https://localhost:8080/\"} id=\"TCMsFBzgfbWA\" outputId=\"1d5fb997-56b1-43c5-8a9f-4e6c4b5112ec\"\n# Получаю батч запросов за июнь, которые были назначены Данилом.\nJQL = 'project = HELPDESK AND created >= 2021-06-01 AND created <= 2021-07-01 AND status in (Решена, Закрыта) AND assignee changed from usc AND assignee changed by (Yajzler.DA) ORDER BY created DESC'\nfields = 'creator, customfield_13590, summary, description, assignee, component'\nhelpdesk_issues = []\nhelpdesk_issues.extend(jira.search_issues(JQL, fields=fields, maxResults=10000))\n\nprint('выгржено запросов: %s' % len(helpdesk_issues))\n\n# + id=\"j1BdJoCshA4Q\"\n# Закрываю сессию Jira\njira.close()\n\n# + colab={\"base_uri\": \"https://localhost:8080/\", \"height\": 450} id=\"HochFmAdgX3x\" outputId=\"f36b7718-62c8-486b-8edd-87b42c64cecd\"\n# Перевожу загруженные данные в датафрейм\nissue_num = []\nsource = []\ncreator = []\nsubject\t= []\nbody = []\nassignee = []\n\nfor iss in helpdesk_issues:\n issue_num.append(iss.key)\n try:\n source.append(iss.fields.customfield_13590)\n except AttributeError:\n source.append(np.nan)\n creator.append(iss.fields.creator.emailAddress)\n subject.append(iss.fields.summary)\n body.append(iss.fields.description)\n try:\n assignee.append(iss.fields.assignee.key)\n except AttributeError:\n assignee.append(np.nan)\n\nfind_assignee = pd.DataFrame(\n {'source':source, 'from':creator, 'subject':subject, 'body':body, 'assignee':assignee},\n index=pd.Index(issue_num, name='issue_number'))\n \nfind_assignee\n\n\n# + [markdown] id=\"A5M2kD5kgQyc\"\n# ### Предобработка запросов\n\n# + colab={\"base_uri\": \"https://localhost:8080/\"} id=\"sSI_eZ1ZkvkX\" outputId=\"426c2265-80eb-46c9-f3f1-21b3342200ee\"\n# Создать словарь\ndef build_vocab(tokenized_texts, \n max_words=20_000, \n min_words_count=3, \n add_tech_tokens=False):\n '''Умеет оставлять в словаре только max_words самых популярных слов. \n min_words_count убирает редкие слова'''\n\n all_words = [word for tokens in tokenized_texts for word in tokens]\n vocab = {}\n for word in all_words:\n if word not in vocab:\n vocab[word] = 1\n else:\n vocab[word] += 1\n vocab = pd.DataFrame({'keys':vocab.keys(), 'values':vocab.values()})\n\n #Добавляю технические слова\n if isinstance(add_tech_tokens, (np.ndarray, list)):\n keys = add_tech_tokens\n values = list(range(999999,999999-len(add_tech_tokens), -1))\n special_words = pd.DataFrame({'keys':keys, 'values':values})\n vocab = pd.concat([vocab, special_words])\n elif add_tech_tokens == 'def':\n special_words = pd.DataFrame({'keys':[\" Exploratory Data Analysis\n\n# + colab={\"base_uri\": \"https://localhost:8080/\"} id=\"Iet2RZX4uMGg\" outputId=\"985763e5-7953-4730-8468-c3287d4efbcd\"\n#observe the different feature type present in the data\nprint(df.dtypes)\nprint(df.info())\n\n# + colab={\"base_uri\": \"https://localhost:8080/\", \"height\": 270} id=\"zFzpjlLA-Xr_\" outputId=\"03df72eb-a8dd-4703-dbef-e21d157327a5\"\ndf.head()\n\n# + colab={\"base_uri\": \"https://localhost:8080/\", \"height\": 1000} id=\"lFiuRmdJuRr6\" outputId=\"a0211aea-0f35-4eb4-f5f5-2899ac2bfef2\"\n# Checking for the missing value present in each columns\ntotal = df.isnull().sum().sort_values(ascending = False)\npercent = (df.isnull().sum()/df.isnull().count()*100).sort_values(ascending = False)\npd.concat([total, percent], axis=1, keys=['Total', 'Percent'])\n\n# + [markdown] id=\"7_OSbYQUuiUy\"\n#
Observation\n#\n# We can see that there is no missing value present in the dataframe.\n#\n#
Outliers treatment\n#\n# PCA transformed data hence it can be safely that outliers are already treated.\n\n# + [markdown] id=\"ynUq5cGB-Xr_\"\n# Here we will observe the distribution of our classes\n\n# + id=\"Z8qEoUkG-Xr_\" colab={\"base_uri\": \"https://localhost:8080/\"} outputId=\"634da353-4031-4933-ebbb-38c5a6179afc\"\nclasses = df['Class'].value_counts()\nnormal_share = classes[0]/df['Class'].count()*100\nfraud_share = classes[1]/df['Class'].count()*100\nnormal_share_num=classes[0]/df['Class'].count()\nfraud_share_num=classes[1]/df['Class'].count()\nprint(\"Normal Share: \", round(normal_share,2))\nprint(\"Fraud Share: \", round(fraud_share,2))\n\n# + [markdown] id=\"p-aVtLyVhOkJ\"\n#
High-Level Observation\n#\n# 1. Highly imbalenced data
XGBoost on balanced data with SMOTE\n\n# + colab={\"base_uri\": \"https://localhost:8080/\", \"height\": 135} id=\"7LXtMqqHS2O9\" outputId=\"5cf1850e-939d-4b97-f38a-0c891f5ca12e\"\n# creating a KFold object\nfolds = 3\n\n# specify range of hyperparameters\nparam_grid = {'learning_rate': [0.2, 0.6],\n 'subsample': [0.3, 0.6, 0.9]}\n\n\n# specify model\nxgb_model = XGBClassifier(max_depth=2, n_estimators=200)\n\n# set up GridSearchCV()\nmodel_cv = GridSearchCV(estimator = xgb_model,\n param_grid = param_grid,\n scoring= 'roc_auc',\n cv = folds,\n verbose = 1,\n return_train_score=True)\n\n# fit the model\nmodel_cv.fit(X_train_smote, y_train_smote)\n\n# + colab={\"base_uri\": \"https://localhost:8080/\", \"height\": 614} id=\"mCgbzHILS2MK\" outputId=\"614e7e5f-c677-4513-b47a-027ce96075c9\"\n# cv results\ncv_results = pd.DataFrame(model_cv.cv_results_)\ncv_results\n\n# + colab={\"base_uri\": \"https://localhost:8080/\"} id=\"wh_p37GnS2Jh\" outputId=\"8576e157-8075-4cff-db4b-c492b20ea695\"\n# Printing the optimal score and hyperparameters\nprint(\"Best roc auc score : \", model_cv.best_score_)\nprint(model_cv.best_estimator_)\n\n# + colab={\"base_uri\": \"https://localhost:8080/\"} id=\"RmE00eU3S2HH\" outputId=\"4113f0d5-0195-46ce-c5dc-7b0f222be6f3\"\nmodel_cv.best_params_\n\n# + colab={\"base_uri\": \"https://localhost:8080/\", \"height\": 304} id=\"9e6tfFF5S2D-\" outputId=\"145ae965-915d-41c6-8450-65d5a8ad3017\"\n# chosen hyperparameters\n# 'objective':'binary:logistic' outputs probability rather than label, which we need for calculating auc\nparams = {'learning_rate': 0.6,\n 'max_depth': 2,\n 'n_estimators':200,\n 'subsample':0.6,\n 'objective':'binary:logistic'}\n\n# fit model on training data\nxgb_smote_model = XGBClassifier(params = params)\nxgb_smote_model.fit(X_train_smote, y_train_smote)\n\n\n# + colab={\"base_uri\": \"https://localhost:8080/\", \"height\": 252} id=\"_h0ttLdSS2Bd\" outputId=\"f8d20785-ac44-4509-e303-50b07cedad85\"\nXGBClassifier(base_score=0.5, booster='gbtree', colsample_bylevel=1,\n colsample_bynode=1, colsample_bytree=1, gamma=0, gpu_id=-1,\n importance_type='gain', interaction_constraints='',\n learning_rate=0.300000012, max_delta_step=0, max_depth=6,\n min_child_weight=1, missing=nan, monotone_constraints='()',\n n_estimators=100, n_jobs=0, num_parallel_tree=1,\n params={'learning_rate': 0.6, 'max_depth': 2, 'n_estimators': 200,\n 'objective': 'binary:logistic', 'subsample': 0.6},\n random_state=0, reg_alpha=0, reg_lambda=1, scale_pos_weight=1,\n subsample=1, tree_method='exact', validate_parameters=1,\n verbosity=None)\n\n# + colab={\"base_uri\": \"https://localhost:8080/\"} id=\"F-_pbwyTS1-3\" outputId=\"bb93e3e9-a0db-4601-e966-15ea7703fa76\"\ny_train_pred = xgb_smote_model.predict(X_train_smote)\ndisplay_scores(y_train_smote, y_train_pred)\n\n# + colab={\"base_uri\": \"https://localhost:8080/\", \"height\": 487} id=\"bICKt7MSS18b\" outputId=\"61c178a5-2703-4aa2-92b1-5532899a14c7\"\n# Predicted probability\ny_train_pred_proba = xgb_smote_model.predict_proba(X_train_smote)[:,1]\n# Plot the ROC curve\ndraw_roc(y_train_smote, y_train_pred_proba)\n\n# + colab={\"base_uri\": \"https://localhost:8080/\"} id=\"FivAHKgAS15n\" outputId=\"adc5c9e7-a2b8-4180-fab0-129b0a3f91c4\"\ny_pred = xgb_smote_model.predict(X_test)\ndisplay_scores(y_test, y_pred)\n\n# + [markdown] id=\"40y8hBqCVI5i\"\n#\n#\n# Model Summary\n#\n# ### Train set\n# - ROC score : 100.0%\n# - F1 score: 100.0%\n# ### Test set\n# - ROC score : 99%\n# - F1 score: 80.95%\n\n# + [markdown] id=\"pUyYD1YWVR7y\"\n# ADASYN (Adaptive Synthetic Sampling)\n\n# + colab={\"base_uri\": \"https://localhost:8080/\", \"height\": 1000} id=\"l9QCqCZES10s\" outputId=\"bbc76fce-9fbb-4299-9a12-ec35c4d3b715\"\nimport warnings\nwarnings.filterwarnings(\"ignore\")\n\nfrom imblearn import over_sampling\n\nada = over_sampling.ADASYN(random_state=0)\nX_train_adasyn, y_train_adasyn = ada.fit_resample(X_train, y_train)\n# Artificial minority samples and corresponding minority labels from ADASYN are appended\n# below X_train and y_train respectively\n# So to exclusively get the artificial minority samples from ADASYN, we do\nX_train_adasyn_1 = X_train_adasyn[X_train.shape[0]:]\n\nX_train_1 = X_train.to_numpy()[np.where(y_train==1.0)]\nX_train_0 = X_train.to_numpy()[np.where(y_train==0.0)]\n\n\n\nimport matplotlib.pyplot as plt\n# %matplotlib inline\nplt.rcParams['figure.figsize'] = [20, 20]\nfig = plt.figure()\n\nplt.subplot(3, 1, 1)\nplt.scatter(X_train_1[:, 0], X_train_1[:, 1], label='Actual Class-1 Examples')\nplt.legend()\n\nplt.subplot(3, 1, 2)\nplt.scatter(X_train_1[:, 0], X_train_1[:, 1], label='Actual Class-1 Examples')\nplt.scatter(X_train_adasyn_1.iloc[:X_train_1.shape[0], 0], X_train_adasyn_1.iloc[:X_train_1.shape[0], 1],\n label='Artificial ADASYN Class-1 Examples')\nplt.legend()\n\nplt.subplot(3, 1, 3)\nplt.scatter(X_train_1[:, 0], X_train_1[:, 1], label='Actual Class-1 Examples')\nplt.scatter(X_train_0[:X_train_1.shape[0], 0], X_train_0[:X_train_1.shape[0], 1], label='Actual Class-0 Examples')\nplt.legend()\n\n# + [markdown] id=\"0Nc3-DrRtjuY\"\n# Logistic Regression on balanced data with ADASYN\n\n# + colab={\"base_uri\": \"https://localhost:8080/\"} id=\"wljXdYUzS1xx\" outputId=\"ed861a4a-a9ca-4588-a17a-a11b54846845\"\n# Creating KFold object with 3 splits\nfolds = KFold(n_splits=3, shuffle=True, random_state=4)\n\n# Specify params\nparams = {\"C\": [0.01, 0.1, 1, 10, 100, 1000]}\n\n# Specifing score as roc-auc\nmodel_cv = GridSearchCV(estimator = LogisticRegression(),\n param_grid = params,\n scoring= 'roc_auc',\n cv = folds,\n verbose = 1,\n return_train_score=True)\n\n# Fit the model\nmodel_cv.fit(X_train_adasyn, y_train_adasyn)\n#print the evaluation result by choosing a evaluation metric\nprint('Best ROC AUC score: ', model_cv.best_score_)\n#print the optimum value of hyperparameters\nprint('Best hyperparameters: ', model_cv.best_params_)\n\n# + colab={\"base_uri\": \"https://localhost:8080/\", \"height\": 371} id=\"fmwV2kLeS1vb\" outputId=\"8e43805e-773c-456f-c2da-e62cdf0dff4e\"\n# cross validation results\ncv_results = pd.DataFrame(model_cv.cv_results_)\ncv_results\n\n# + colab={\"base_uri\": \"https://localhost:8080/\", \"height\": 546} id=\"PszAYhxLS1pP\" outputId=\"ce1fd241-69df-423d-8e08-5d2042083668\"\n# plot of C versus train and validation scores\n\nplt.figure(figsize=(8, 6))\nplt.plot(cv_results['param_C'], cv_results['mean_test_score'])\nplt.plot(cv_results['param_C'], cv_results['mean_train_score'])\nplt.xlabel('C')\nplt.ylabel('sensitivity')\nplt.legend(['test result', 'train result'], loc='upper left')\nplt.xscale('log')\n\n# + colab={\"base_uri\": \"https://localhost:8080/\"} id=\"jeRY9KZ1S1lF\" outputId=\"661a054e-fb92-440f-8741-1c23c6ca76ca\"\nmodel_cv.best_params_\n\n# + colab={\"base_uri\": \"https://localhost:8080/\", \"height\": 75} id=\"aKOaj6oiS1ib\" outputId=\"70edccc6-2cae-4068-fef8-3c72a6738d8c\"\n# Instantiating the model\nlogreg_adasyn_model = LogisticRegression(C=1000)\n\n# Fitting the model\nlogreg_adasyn_model.fit(X_train_adasyn, y_train_adasyn)\n\n# + colab={\"base_uri\": \"https://localhost:8080/\"} id=\"emAQrFosS1fU\" outputId=\"e571a041-3d2a-4864-8f7e-7598c41ef5b0\"\n# Evaluating on test data\ny_train_pred = logreg_adasyn_model.predict(X_train_adasyn)\ndisplay_scores(y_train_adasyn, y_train_pred)\n\n# + colab={\"base_uri\": \"https://localhost:8080/\", \"height\": 487} id=\"bI5Ecx6WS1cW\" outputId=\"6354a94d-95e1-4bf0-b56f-c97f813d1b21\"\n# Predicted probability\ny_train_pred_proba = logreg_adasyn_model.predict_proba(X_train_adasyn)[:,1]\n# Plot the ROC curve\ndraw_roc(y_train_adasyn, y_train_pred_proba)\n\n# + colab={\"base_uri\": \"https://localhost:8080/\"} id=\"lV51N0pTS1YR\" outputId=\"9290e953-23b1-42f2-9259-ad53dde7d77c\"\n# Evaluating on test data\ny_pred = logreg_adasyn_model.predict(X_test)\ndisplay_scores(y_test, y_pred)\n\n# + colab={\"base_uri\": \"https://localhost:8080/\", \"height\": 487} id=\"U6VCtxjRS1Vo\" outputId=\"840c37d2-2974-4c07-9862-33c5dc335501\"\n# Predicted probability\ny_test_pred_proba = logreg_adasyn_model.predict_proba(X_test)[:,1]\n# Plot the ROC curve\ndraw_roc(y_test, y_test_pred_proba)\n\n# + [markdown] id=\"7GSC-V2MupHD\"\n# Model Summary\n#\n# ### Train set\n# - ROC score : 96%\n# - F1 score: 89.2%\n# ### Test set\n# - ROC score : 98%\n# - F1 score: 3.23%\n\n# + [markdown] id=\"CM04wZANupqS\"\n# Decision Tree on balanced data with ADASYN\n#\n\n# + colab={\"base_uri\": \"https://localhost:8080/\", \"height\": 135} id=\"eYH1CmcwS1So\" outputId=\"cd713506-6cb2-4a25-d6e8-10c0a60013e0\"\n# Create the parameter grid\nparam_grid = {\n 'max_depth': range(5, 15, 5),\n 'min_samples_leaf': range(50, 150, 50),\n 'min_samples_split': range(50, 150, 50),\n}\n\n\n# Instantiate the grid search model\ndtree = DecisionTreeClassifier()\n\ngrid_search = GridSearchCV(estimator = dtree,\n param_grid = param_grid,\n scoring= 'roc_auc',\n cv = 5,\n n_jobs=-1,\n verbose = 1)\n\n# Fit the grid search to the data\ngrid_search.fit(X_train_adasyn,y_train_adasyn)\n\n# + colab={\"base_uri\": \"https://localhost:8080/\"} id=\"j9vAr8xPujuP\" outputId=\"9ab0a4b0-9043-4ba4-d1d7-a8ee83112c0f\"\n# Printing the optimal roc score and hyperparameters\nprint(\"Best roc auc score : \", grid_search.best_score_)\nprint(grid_search.best_estimator_)\n\n# + colab={\"base_uri\": \"https://localhost:8080/\", \"height\": 92} id=\"Hu6w3VRzujqu\" outputId=\"ad40660c-bb8e-4fdb-eb97-c6ce99fd8616\"\n# Model with optimal hyperparameters\ndt_adasyn_model = DecisionTreeClassifier(criterion = \"gini\",\n random_state = 100,\n max_depth=10,\n min_samples_leaf=50,\n min_samples_split=100)\n\ndt_adasyn_model.fit(X_train_adasyn, y_train_adasyn)\n\n# + colab={\"base_uri\": \"https://localhost:8080/\"} id=\"ZBkR5KuGujnR\" outputId=\"f3c33918-01ef-4091-e632-4a6ab4313436\"\n# Evaluating model on the test data\ny_train_pred = dt_adasyn_model.predict(X_train_adasyn)\ndisplay_scores(y_train_adasyn, y_train_pred)\n\n# + colab={\"base_uri\": \"https://localhost:8080/\", \"height\": 487} id=\"PfGFsiT8ujko\" outputId=\"756769f0-f5c0-4e3f-9bff-20e96f82de92\"\n# Predicted probability\ny_train_pred_proba = dt_adasyn_model.predict_proba(X_train_adasyn)[:,1]\n# Plot the ROC curve\ndraw_roc(y_train_adasyn, y_train_pred_proba)\n\n# + colab={\"base_uri\": \"https://localhost:8080/\"} id=\"I9vNQZFMujhP\" outputId=\"e6cd0b6c-28b3-415b-dfea-0bf9939ddce5\"\n# Evaluating model on the test data\ny_pred = dt_adasyn_model.predict(X_test)\ndisplay_scores(y_test, y_pred)\n\n# + colab={\"base_uri\": \"https://localhost:8080/\", \"height\": 487} id=\"hQfuzGIUujeR\" outputId=\"8dca4431-b9f1-436a-993d-1dd82cbced80\"\n# Predicted probability\ny_test_pred_proba = dt_adasyn_model.predict_proba(X_test)[:,1]\n# Plot the ROC curve\ndraw_roc(y_test, y_test_pred_proba)\n\n# + [markdown] id=\"L3pOZ3bCvdfi\"\n# Model Summary\n#\n# ### Train set\n# - ROC score : 99%\n# - F1 score: 97.79%\n#\n# ### Test set\n# - ROC score : 95%\n# - F1 score: 7.05%\n\n# + [markdown] id=\"xVAiRVxMvdcj\"\n# XGBoost on balanced data with ADASYN\n\n# + colab={\"base_uri\": \"https://localhost:8080/\", \"height\": 135} id=\"j64gH15lvdEQ\" outputId=\"ad34a1f9-7ef2-4680-def0-ccbba9a7dc9c\"\n# creating a KFold object\nfolds = 3\n\n# specify range of hyperparameters\nparam_grid = {'learning_rate': [0.2, 0.6],\n 'subsample': [0.3, 0.6, 0.9]}\n\n\n# specify model\nxgb_model = XGBClassifier(max_depth=2, n_estimators=200)\n\n# set up GridSearchCV()\nmodel_cv = GridSearchCV(estimator = xgb_model,\n param_grid = param_grid,\n scoring= 'roc_auc',\n cv = folds,\n verbose = 1,\n return_train_score=True)\n\n# fit the model\nmodel_cv.fit(X_train_adasyn, y_train_adasyn)\n\n# + colab={\"base_uri\": \"https://localhost:8080/\", \"height\": 614} id=\"zlJxz5BEvdBP\" outputId=\"c0266621-c007-4878-ca2f-250aa7ae47f6\"\n# cv results\ncv_results = pd.DataFrame(model_cv.cv_results_)\ncv_results\n\n# + colab={\"base_uri\": \"https://localhost:8080/\"} id=\"7jpFD8sdujbY\" outputId=\"ff26b563-6c4d-4b52-c4d5-24f1f0ec3902\"\n# Printing the optimal score and hyperparameters\nprint(\"Best roc auc score : \", model_cv.best_score_)\nprint(model_cv.best_estimator_)\n\n# + colab={\"base_uri\": \"https://localhost:8080/\"} id=\"RxPtYR2LujYo\" outputId=\"7700b982-eab8-45dc-8900-c14ee48b2680\"\nmodel_cv.best_params_\n\n# + colab={\"base_uri\": \"https://localhost:8080/\", \"height\": 304} id=\"g8oYHUbUujWn\" outputId=\"04200f72-cb32-4789-b444-1b20e580b34c\"\n# chosen hyperparameters\n# 'objective':'binary:logistic' outputs probability rather than label, which we need for calculating auc\nparams = {'learning_rate': 0.6,\n 'max_depth': 2,\n 'n_estimators':200,\n 'subsample':0.9,\n 'objective':'binary:logistic'}\n\n# Model with optimal hyperparameter\nxgb_adasyn_model = XGBClassifier(params = params)\nxgb_adasyn_model.fit(X_train_adasyn,y_train_adasyn)\n\n# + colab={\"base_uri\": \"https://localhost:8080/\"} id=\"E00h5FDEujTN\" outputId=\"dfc63e5f-c0b5-4d00-b586-0c45c3108c44\"\n# Predicting on the train set\ny_train_pred = xgb_adasyn_model.predict(X_train_adasyn)\n# Printing the scores\ndisplay_scores(y_train_adasyn, y_train_pred)\n\n# + colab={\"base_uri\": \"https://localhost:8080/\", \"height\": 487} id=\"flObdb0oujP3\" outputId=\"1a8e7cb6-5c9c-4fd9-9e11-f36953ebf229\"\n# Predicted probability\ny_train_pred_proba = xgb_adasyn_model.predict_proba(X_train_adasyn)[:,1]\n# Plot the ROC curve\ndraw_roc(y_train_adasyn, y_train_pred_proba)\n\n# + colab={\"base_uri\": \"https://localhost:8080/\"} id=\"vxymv_crujNH\" outputId=\"03f36655-62ac-44a7-8707-1716f814365c\"\ny_pred = xgb_adasyn_model.predict(X_test)\ndisplay_scores(y_test, y_pred)\n\n# + colab={\"base_uri\": \"https://localhost:8080/\", \"height\": 487} id=\"M_w-oPWEujJ_\" outputId=\"6c5d1985-4a61-45e8-8c80-b47b552b3bde\"\n# Predicted probability\ny_test_pred_proba = xgb_adasyn_model.predict_proba(X_test)[:,1]\n# Plot the ROC curve\ndraw_roc(y_test, y_test_pred_proba)\n\n# + [markdown] id=\"qJFxSFumwaRX\"\n# Model Summary\n#\n# ### Train set\n# - ROC score : 100.0%\n# - F1 score: 100.0%\n#\n# Test set\n# - ROC score : 99%\n# - F1 score: 81.16%\n\n# + [markdown] id=\"9O-WsCFHwfqZ\"\n# Select the best result on a model
\n\n# + [markdown] id=\"ZWjnfM84rNQ1\"\n# # Heart Diease Classification\n\n# + [markdown] id=\"SfHnOsj_rmuk\"\n# ## Import\n\n# + id=\"lQvgmcl5rmdp\"\nimport pandas as pd\nimport numpy as np\nimport matplotlib.pyplot as plt\nimport seaborn as sns\nimport tensorflow as tf\nfrom tensorflow import keras\n#from timeit import default_timer as timer\n#from datetime import timedelta\n\n# + id=\"vfvi-lRJrJzi\"\nheart = pd.read_csv(\"/content/heart.csv\")\n\n# + [markdown] id=\"7ieuY_Vfr8Nn\"\n# * Age : Age of the patient 1: male, 0: female\n# * Sex : Sex of the patient\n# * Lexang: exercise induced angina (1 = yes; 0 = no)\n# * ca: number of major vessels (0-3)\n#\n# * cp : Chest Pain type chest pain type\n# Value 1: typical angina\n# Value 2: atypical angina\n# Value 3: non-anginal pain\n# Value 4: asymptomatic\n# * trtbps : resting blood pressure (in mm Hg)\n# * chol : cholestoral in mg/dl fetched via BMI sensor\n# * fbs : (fasting blood sugar > 120 mg/dl) (1 = true; 0 = false)\n# * rest_ecg : resting electrocardiographic results\n# Value 0: normal\n# Value 1: having ST-T wave abnormality (T wave inversions and/or ST elevation or depression of > 0.05 mV)\n# Value 2: showing probable or definite left ventricular hypertrophy by Estes' criteria\n# * thalach : maximum heart rate achieved\n# * target : 0= less chance of heart attack 1= more chance of heart attack\n\n# + colab={\"base_uri\": \"https://localhost:8080/\", \"height\": 204} id=\"Selo6iJfrhy8\" outputId=\"4677fc84-5189-45d5-d20b-a352f77fb26a\"\nheart.head()\n\n# + colab={\"base_uri\": \"https://localhost:8080/\"} id=\"OuKEJ6lFsyYX\" outputId=\"91a3e3e9-381c-4448-caa5-7a30e9e24076\"\nheart.info()\n\n# + colab={\"base_uri\": \"https://localhost:8080/\"} id=\"GTLh9FgVrhpU\" outputId=\"1adc346e-a72b-45c7-ce0b-9e9df7951b75\"\nfor i in heart.columns:\n print(i,\"unique value:\")\n print(\"=>\", heart[i].unique())\n\n# + id=\"FVIE_HNlrher\"\ncategorical_columns = [\"sex\",\"cp\",\"fbs\",\"restecg\",\"exang\",\"slope\",\"ca\",\"thal\"]\n\n# + [markdown] id=\"63tDSp0IriS-\"\n# ## EDA\n\n# + id=\"PTQAJAsZC4_-\"\nEDA = heart.copy()\n\n# + id=\"RQqBv0EYAQQv\"\nEDA.sex = EDA.sex.astype('str')\n\n# + colab={\"base_uri\": \"https://localhost:8080/\"} id=\"8yi2SWWI_b7a\" outputId=\"c1ae5000-5629-4e5b-8da8-81480dbf8414\"\nEDA.sex[EDA.sex == '0'] = 'female'\nEDA.sex[EDA.sex == '1'] = 'male'\n\n# + colab={\"base_uri\": \"https://localhost:8080/\"} id=\"cZdFZN238rfC\" outputId=\"c2ce0b67-8737-4174-99e8-82694a850267\"\nEDA.columns\n\n# + colab={\"base_uri\": \"https://localhost:8080/\", \"height\": 920} id=\"MUJxMgoi6u2r\" outputId=\"deeee2cd-890e-40d7-8d3b-9290d81d9163\"\nsns.set_style(\"whitegrid\")\nsns.pairplot(EDA[['age', 'trestbps', 'chol', 'thalach', 'oldpeak']])\n\n# + id=\"RSmhYeaX9JQD\"\n\n\n# + colab={\"base_uri\": \"https://localhost:8080/\", \"height\": 510} id=\"4ZDWW1ijxpvr\" outputId=\"78a35e0a-f876-48c4-d446-99ebeee22f9e\"\nfig, ax = plt.subplots(1,3, figsize = (15,7))\n\nsns.histplot(x='age', data =EDA, ax = ax[0])\nsns.histplot(x='age', data =EDA[EDA.sex == 'male'], ax = ax[1])\nsns.histplot(x='age', data =EDA[EDA.sex == 'female'], ax = ax[2])\n\nax[0].set_title('Whole')\nax[1].set_title('Male')\nax[2].set_title('Female')\n\nplt.suptitle(\"Age Distribution\")\n\n# + colab={\"base_uri\": \"https://localhost:8080/\", \"height\": 296} id=\"U7fUHMAG_NtQ\" outputId=\"2efc60e6-0b2d-4015-ae96-52e571d3c5b1\"\nsns.histplot(x = \"sex\", data = EDA)\n\n# + colab={\"base_uri\": \"https://localhost:8080/\", \"height\": 459} id=\"FqOKJqxMndMj\" outputId=\"49f1a328-1ebf-491d-8016-5a9feac229b3\"\nfig, ax = plt.subplots(1,2, figsize = (12,7))\n\nsns.boxplot(x = 'sex', y = 'thalach', data = EDA, ax = ax[0])\nsns.boxplot(x = 'target', y = 'thalach', data = EDA, ax = ax[1])\n\n# + id=\"yhUiTkxhrTap\"\nmale_1_index = (EDA.target == 1) & (EDA.sex == 'male')\nfemale_1_index = (EDA.target == 1) & (EDA.sex == 'female')\n\nmale_0_index = (EDA.target == 0) & (EDA.sex == 'male')\nfemale_0_index = (EDA.target == 0) & (EDA.sex == 'female')\n\n# + id=\"BQe2zbBDsNh2\"\nprop_male_1 = sum(male_1_index)/sum(EDA.sex == 'male')\nprop_female_1 = sum(female_1_index)/sum(EDA.sex == 'female')\n\n# + colab={\"base_uri\": \"https://localhost:8080/\"} id=\"JBsmDNfFshzo\" outputId=\"8217b639-56fd-42af-ee22-1f2120c87879\"\nprint('The percetage of males who have the heart diese',prop_male_1)\nprint('The percetage of females who have the heart diese',prop_female_1)\n\n# + id=\"JJvxhgr0s0Yv\"\nprop_male_0 = sum(male_0_index)/sum(EDA.sex == 'male')\nprop_female_0 = sum(female_0_index)/sum(EDA.sex == 'female')\n\n# + colab={\"base_uri\": \"https://localhost:8080/\"} id=\"2qPb6OEVs6Iz\" outputId=\"8d94bc34-27c5-4324-93ac-e45cb9730ffa\"\n# must be 1\nprint(prop_male_0 + prop_male_1)\nprint(prop_female_0 + prop_female_1)\n\n# + colab={\"base_uri\": \"https://localhost:8080/\", \"height\": 381} id=\"8n5BEGiCtZZ3\" outputId=\"d29dc521-0f6a-4c8e-b7ee-4538ac41d873\"\nfig, ax = plt.subplots(1,2, figsize=(10,6))\n\nax[0].pie(x=[prop_male_0,prop_male_1], autopct=\"%.1f%%\", explode=[0.01]*2, labels=[\"Dieses\",\"No Dieses\"], pctdistance=0.5)\nax[1].pie(x=[prop_female_0,prop_female_1], autopct=\"%.1f%%\", explode=[0.01]*2, labels=[\"Dieses\",\"No Dieses\"], pctdistance=0.5)\n\nax[0].set_title('Male')\nax[1].set_title('Female')\n\nfig.suptitle(\"Percentage\")\n\n# + [markdown] id=\"7LOffSzkwGjj\"\n# ***\n\n# + [markdown] id=\"j80dL4Pyv-OI\"\n# ## Data Engineering\n\n# + id=\"t67Ur6b-tfof\"\ny = heart.target\nX = heart.drop([\"target\"], axis = 1)\n\n# + [markdown] id=\"mLYDE0Fqt_vi\"\n# ### One Hot Encoding\n\n# + id=\"3x8esCcFt_Jm\"\nfrom sklearn.preprocessing import OneHotEncoder\nohe = OneHotEncoder(drop='first')\n\n# + id=\"sszZt7yluE1I\"\nnon_categorical = ['age', 'trestbps', 'chol', 'thalach', 'oldpeak']\n\n# + id=\"Vqm-Z3_RuWCB\"\nohe_dataset = ohe.fit_transform(heart[categorical_columns])\ncolumns_ohe = ohe.get_feature_names(input_features = heart[categorical_columns].columns)\nohe_dataset = pd.DataFrame(ohe_dataset.toarray())\nohe_dataset.columns = columns_ohe\nX_OH = pd.concat([X.drop(categorical_columns,axis = 1),ohe_dataset],axis = 1)\n\n# + colab={\"base_uri\": \"https://localhost:8080/\", \"height\": 204} id=\"U538PrSqtv0t\" outputId=\"6986ae3e-bc65-4c61-d5ee-6c3da6a438bd\"\n# One Hot Encoding\nX_OH.head()\n\n# + [markdown] id=\"PPFlGx_PvKwx\"\n# ### Normalization\n\n# + id=\"hc9fbZinvKEI\"\nfrom sklearn.preprocessing import MinMaxScaler\nscaler = MinMaxScaler()\n\nmin_heart = X_OH.min()\nmax_heart = X_OH.max()\n\nX_normalized = scaler.fit_transform(X_OH)\n\nX_normalized = pd.DataFrame(X_normalized)\n\nX_normalized.columns = X_OH.columns\n\n# + colab={\"base_uri\": \"https://localhost:8080/\", \"height\": 204} id=\"DX9yKXDxvttw\" outputId=\"a189c219-3a06-467d-f8a3-bfb401ea93a0\"\nX_normalized.head()\n\n# + id=\"tCObFAvxKwC4\"\nfrom sklearn.preprocessing import StandardScaler\nscaler = StandardScaler()\n\nmin_heart = X_OH.min()\nmax_heart = X_OH.max()\n\nX_std = scaler.fit_transform(X_OH)\n\nX_std = pd.DataFrame(X_std)\n\nX_std.columns = X_OH.columns\n\n# + colab={\"base_uri\": \"https://localhost:8080/\", \"height\": 204} id=\"pz9skFGoK7vZ\" outputId=\"0fd18c48-e536-4164-9eff-3fcb9deefc9d\"\nX_std.head()\n\n# + [markdown] id=\"Amj6qMO3wMuO\"\n# ### Splitting Dataset\n\n# + id=\"5j4NqfD4wZyD\"\nfrom sklearn.model_selection import train_test_split \n\n# + id=\"ZvG_Blv9wUQi\"\nX_train, X_test, y_train, y_test = train_test_split(\n X_std, \n y, \n test_size=0.2, \n random_state=629) \n\nX_train_tree, X_test_tree, y_train_tree, y_test_tree = train_test_split(\n X, \n y, \n test_size=0.2, \n random_state=629) \n\n# + [markdown] id=\"iUqP1Meec2WN\"\n# # Binary Classifers\n\n# + id=\"JvjVyb4rlx-0\"\nfrom sklearn import metrics\n\n# + [markdown] id=\"8YCoAV-EvG_n\"\n# ## DNN\n\n# + [markdown] id=\"WtbfPgyH-LiA\"\n# The number of examples is 303, which is quite little to feed DNN model. But just for FUN we can give it a try!\n\n# + id=\"C36EppIix2I0\"\nfrom tensorflow.keras import regularizers\n\n# + id=\"vm4noJlEtNVY\"\nval_acc_threshold = 0.93\nacc_threshold = 0.95\n\n\n# + id=\"A6KlhrF_tVUn\"\nclass myCallback(tf.keras.callbacks.Callback): \n def on_epoch_end(self, epoch, logs={}): \n if((logs.get('accuracy') > acc_threshold) & (logs.get('val_accuracy') > val_acc_threshold)): \n print(\"Reached val_accuracy, so stopping training!!\".format(acc_threshold))\n self.model.stop_training = True\n\n\n# + id=\"h_R8oizhuU5e\"\ncallbacks = myCallback()\n\n# + colab={\"base_uri\": \"https://localhost:8080/\"} id=\"FTNngVU3wL5V\" outputId=\"3c1a7c59-87fa-4c23-9c21-abc4af217c3e\"\nmodel = keras.Sequential([\n keras.layers.Dense(256,activation = \"relu\",kernel_regularizer=regularizers.l2(l2=1e-2)),\n keras.layers.BatchNormalization(),\n keras.layers.Dense(128,activation = \"relu\",kernel_regularizer=regularizers.l2(l2=1e-2)),\n keras.layers.BatchNormalization(),\n keras.layers.Dense(64,activation = \"relu\",kernel_regularizer=regularizers.l2(l2=1e-2)),\n keras.layers.BatchNormalization(),\n keras.layers.Dense(16,activation = \"relu\",kernel_regularizer=regularizers.l2(l2=1e-2)),\n keras.layers.BatchNormalization(),\n keras.layers.Dense(1,activation=\"sigmoid\")\n ])\n## Callback\n \n\nadam_opt = keras.optimizers.Adam(lr=0.001)\nmodel.compile(optimizer=adam_opt,loss = 'binary_crossentropy', metrics=['accuracy'])\n\n\nhistory = model.fit(X_train,\n y_train,\n epochs=300,\n verbose=1,\n validation_data = (X_test,y_test),\n callbacks=[callbacks])\n\n# + colab={\"base_uri\": \"https://localhost:8080/\", \"height\": 282} id=\"nzlmht_oobhg\" outputId=\"cc20cbe3-5ab0-48c0-c6dc-0428ddbb4a27\"\nplt.plot(history.history[\"accuracy\"])\nplt.plot(history.history[\"val_accuracy\"])\n\n# + colab={\"base_uri\": \"https://localhost:8080/\"} id=\"QK3V28MlVwoz\" outputId=\"46a256ad-3df8-4508-cc10-b68d021a9031\"\ny_dnn_pred = model.predict_classes(X_test)\n\n# + id=\"9bYqAtRQWEir\"\ny_dnn_pred = np.squeeze(y_dnn_pred)\n\n# + colab={\"base_uri\": \"https://localhost:8080/\"} id=\"c2vjsIJWV_Wm\" outputId=\"185e4ad3-846f-441f-fe47-ab602b8089bc\"\nprint(\"Accuracy:\",metrics.accuracy_score(y_test, y_dnn_pred))\n\n# + [markdown] id=\"2Wa7trnhIemX\"\n# ## Support Vector Machine\n\n# + id=\"Lz53SCT9IeHV\"\nfrom sklearn import svm\nfrom sklearn.model_selection import GridSearchCV\n\n# + id=\"6z0f5oodL07r\"\nparam_grid = {'C': [0.1,1, 10, 100], 'gamma': [1,0.1,0.01,0.001],'kernel': ['rbf', 'poly', 'sigmoid']}\n\n# + colab={\"base_uri\": \"https://localhost:8080/\"} id=\"u7MEg80eL5ku\" outputId=\"1d44f442-5436-4f3a-ab3e-c0931e4f794a\"\ngrid = GridSearchCV(svm.SVC(),param_grid,refit=True,verbose=0)\ngrid.fit(X_train,y_train)\n\n# + id=\"u4P32_9aL_tx\" colab={\"base_uri\": \"https://localhost:8080/\"} outputId=\"decb14cf-8cf0-4813-8123-b6aa87c06893\"\nprint(grid.best_estimator_)\n\n# + id=\"z0s-jSDBMC7K\"\nclf = svm.SVC(C=10, break_ties=False, cache_size=200, class_weight=None, coef0=0.0,\n decision_function_shape='ovr', degree=3, gamma=0.1, kernel='sigmoid',\n max_iter=-1, probability=False, random_state=None, shrinking=True,\n tol=0.001, verbose=False)\n\n# + id=\"zV90gd3jIpDH\" colab={\"base_uri\": \"https://localhost:8080/\"} outputId=\"494e0756-87f3-4a3f-c820-4597f659a594\"\nclf.fit(X_train, y_train)\n\n# + id=\"k-l6rFkgKixZ\"\ny_svm_pred = clf.predict(X_test)\n\n# + id=\"C6ogiJM_KeP-\" colab={\"base_uri\": \"https://localhost:8080/\"} outputId=\"5356447b-baca-4a08-e761-7c7116003546\"\nprint(\"Accuracy:\",metrics.accuracy_score(y_test, y_svm_pred))\n\n# + [markdown] id=\"zQg7KS93MTOl\"\n# ## Random Forest\n\n# + id=\"eRRomm4SMSrH\"\nfrom sklearn.ensemble import RandomForestClassifier\n\n# + id=\"evi3ZjHCM235\"\n#param_grid = {'max_depth': [80, 90, 100, 110],\n# 'min_samples_leaf': [3, 4, 5],\n# 'min_samples_split': [8, 10, 12],\n# 'n_estimators': [100, 200, 300, 1000]}\n\n# + id=\"I2T7UyI9NDU9\"\n#grid = GridSearchCV(RandomForestClassifier(),param_grid,refit=True,verbose=2)\n#grid.fit(X,y)\n\n# + id=\"AoFHgLC8NZ8o\"\n#print(grid.best_estimator_)\n\n# + id=\"Im-bncEOQQTl\"\nclf = RandomForestClassifier()\n\n# + id=\"3tsRlJ-UMi5T\" colab={\"base_uri\": \"https://localhost:8080/\"} outputId=\"76798230-0d3f-4610-8de4-f6aca7c3ec43\"\nclf.fit(X_train_tree, y_train_tree)\n\n# + id=\"N-bjDHn4MitE\"\ny_rf_pred = clf.predict(X_test_tree)\n\n# + id=\"Wd04W1mkMiav\" colab={\"base_uri\": \"https://localhost:8080/\"} outputId=\"41067c3f-8c5a-43ea-cdec-bee0a27b225c\"\nprint(\"Accuracy:\",metrics.accuracy_score(y_test_tree, y_rf_pred))\n\n# + [markdown] id=\"CjGqlBftQguH\"\n# ## Adaboost\n\n# + id=\"bjdy88eqQAua\"\nfrom sklearn.ensemble import AdaBoostClassifier\n\n# + id=\"Tkz3SOurQjEz\"\nclf = AdaBoostClassifier(n_estimators=100, random_state=0)\n\n# + id=\"iUTqXNllQnNa\" colab={\"base_uri\": \"https://localhost:8080/\"} outputId=\"41b63a12-05e1-4962-ee7b-1e505ef1ab2c\"\nclf.fit(X_train, y_train)\n\n# + id=\"Op8vbR8HQo3c\"\ny_ada_pred = clf.predict(X_test)\n\n# + id=\"_iKFEPDcQrQy\" colab={\"base_uri\": \"https://localhost:8080/\"} outputId=\"ffdf074d-5a2b-49d6-a525-bf52d9dd51d3\"\nprint(\"Accuracy:\",metrics.accuracy_score(y_test, y_ada_pred))\n\n# + [markdown] id=\"Lm3ViWysRIJD\"\n# ## Gradient Boosting\n\n# + id=\"iZuWF0cuQtyc\"\nfrom sklearn.ensemble import GradientBoostingClassifier\n\n# + id=\"qcMS4uKiRNYS\"\nclf = GradientBoostingClassifier(n_estimators=100, learning_rate=1.0, max_depth=1, random_state=0)\n\n# + id=\"kGbz0JDaRfDN\" colab={\"base_uri\": \"https://localhost:8080/\"} outputId=\"0fda0d41-9900-440c-a328-7cf94bce349c\"\nclf.fit(X_train, y_train)\n\n# + id=\"XSOuwL7cRi_i\"\ny_gb_pred = clf.predict(X_test)\n\n# + id=\"HLxcWvjNRmW2\" colab={\"base_uri\": \"https://localhost:8080/\"} outputId=\"91c75650-f2a8-42aa-c002-a6ec9887af63\"\nprint(\"Accuracy:\",metrics.accuracy_score(y_test, y_gb_pred))\n\n# + [markdown] id=\"MCRUJZFtS7UU\"\n# ## Logistic Regression\n\n# + id=\"hL5kfLWsSejl\"\nfrom sklearn.linear_model import LogisticRegression\n\n# + id=\"GS9hVIxOS-uA\"\nclf = LogisticRegression(random_state=0)\n\n# + id=\"QnYSI7aATCCP\" colab={\"base_uri\": \"https://localhost:8080/\"} outputId=\"6384b447-cebe-497e-f660-e666b761292c\"\nclf.fit(X_train, y_train)\n\n# + id=\"FU5Po1UXTTLC\"\ny_logistic_pred = clf.predict(X_test)\n\n# + id=\"MSbt8pmITnQC\" colab={\"base_uri\": \"https://localhost:8080/\"} outputId=\"5b76e2c4-be73-4d6e-b04b-554e94fc9ccb\"\nprint(\"Accuracy:\",metrics.accuracy_score(y_test, y_logistic_pred))\n\n# + id=\"fjJzb3RVdgUi\"\n# To Do List\nfrom sklearn.neighbors import KNeighborsClassifier\nfrom sklearn.naive_bayes import BernoulliNB\nfrom sklearn.naive_bayes import GaussianNB\n\n# + id=\"43Jsecx-aSkd\"\nfrom sklearn.metrics import confusion_matrix\n\n# + [markdown] id=\"YsfElX47gIBf\"\n# ## Gaussian Naive Bayes\n\n# + id=\"qSGOqMG5d8dc\" colab={\"base_uri\": \"https://localhost:8080/\"} outputId=\"f4cfea10-1031-4d20-bdec-e875cc062a29\"\nclf = GaussianNB()\nclf.fit(X_train, y_train)\n \ny_Gaussian_pred = clf.predict(X_test)\n\nprint(\"Accuracy of Gaussian Naive Bayes:\", round(metrics.accuracy_score(y_test, y_Gaussian_pred),3))\n\n# + [markdown] id=\"ch3H79akgOFZ\"\n# ## Bernoulli Naive Bayes\n\n# + id=\"2p7JeK3yemAp\" colab={\"base_uri\": \"https://localhost:8080/\"} outputId=\"ffe75c0d-923d-4c42-b4aa-4e34be3d55c3\"\nclf = BernoulliNB()\nclf.fit(X_train, y_train)\n \ny_Bernoulli_pred = clf.predict(X_test)\n \nprint(\"The accuracy of Bernoulli Naive Bayes: \", round(metrics.accuracy_score(y_test, y_Bernoulli_pred),3))\n\n# + [markdown] id=\"1SSaWiExggji\"\n# ## K Nearest Neighbours\n\n# + id=\"FTzdhiGVfP_n\" colab={\"base_uri\": \"https://localhost:8080/\"} outputId=\"9f745f61-331e-4a58-ecf0-74c841c68ffc\"\nclf = KNeighborsClassifier(n_neighbors = 1) \nclf.fit(X_train, y_train)\n\ny_KNeighbors_pred = model.predict(X_test).round()\n\nprint(\"The accuracy of K Nearest Neighbours : \", round(metrics.accuracy_score(y_test, y_KNeighbors_pred),3))\n\n# + [markdown] id=\"Dmyay26bT0X_\"\n# # Summary\n\n# + id=\"-qHASpJ7TzQ5\" colab={\"base_uri\": \"https://localhost:8080/\"} outputId=\"618fff65-0923-4fd7-812f-640d982914ad\"\nnames = [\"Deep Neural Network\", \"Support Vector Machine\", \"Random Forest\", \"Adaboost\",\"Gradient Boost\",\"Logistic Regression\",\n \"Gaussian Naive Bayes\", \"Bernoulli Naive Bayes\",\"K Nearest Neighbours\"]\n\npredictions = [y_dnn_pred, y_svm_pred, y_rf_pred, y_ada_pred, y_gb_pred, y_logistic_pred, y_Gaussian_pred,\n y_Bernoulli_pred, y_KNeighbors_pred]\n\nacc_list = []\nfor i in range(len(names)):\n score = metrics.accuracy_score(y_test,predictions[i])\n print(\"Accuracy of\",names[i], \":\", round(score,3))\n acc_list.append(score)\n\nindex_max = np.argmax(acc_list)\nprint(\"\\n\")\nprint(\"The best model is << {0} >> with accuracy:{1}\".format(names[index_max],round(acc_list[index_max],3)))\n\n# + id=\"AlbgzK1Yfp8C\"\n#error_rate = []\n# \n#for i in range(1, 40):\n# \n# model = KNeighborsClassifier(n_neighbors = i)\n# model.fit(x_train, y_train)\n# pred_i = model.predict(x_test)\n# error_rate.append(np.mean(pred_i != y_test))\n \n#plt.figure(figsize =(10, 6))\n#plt.plot(range(1, 40), error_rate, color ='blue',\n# linestyle ='dashed', marker ='o',\n# markerfacecolor ='red', markersize = 10)\n \n#plt.title('Error Rate vs. K Value')\n#plt.xlabel('K')\n#plt.ylabel('Error Rate')\n","repo_name":"TaeminDA/portfolio","sub_path":"basic/binary/Heart_Diease_Classification.ipynb","file_name":"Heart_Diease_Classification.ipynb","file_ext":"py","file_size_in_byte":17118,"program_lang":"python","lang":"en","doc_type":"code","stars":0,"dataset":"github-jupyter-script","pt":"44"}
+{"seq_id":"28854980061","text":"# # Importing All Libraries\n\n# +\nimport pandas as pd\nimport numpy as np\nimport matplotlib.pyplot as plt\n\nfrom tensorflow.keras.models import Sequential\nfrom tensorflow.keras.layers import Dense, BatchNormalization, Dropout, Activation\nfrom tensorflow.keras.layers import LSTM, GRU\nfrom sklearn.metrics import mean_squared_error\nfrom tensorflow.keras.optimizers import Adam, SGD, RMSprop\nfrom sklearn.metrics import mean_squared_error\nfrom tensorflow.keras.optimizers import Adam, SGD, RMSprop\n\nimport tensorflow as tf\nfrom tensorflow.keras.layers import Input, Conv2D, Dense, Flatten, Dropout, BatchNormalization,LSTM,Bidirectional,MaxPooling2D,GlobalMaxPooling2D,TimeDistributed\nfrom tensorflow.keras.models import Model, Sequential\nfrom tensorflow.keras.optimizers import Adam\nfrom tensorflow.keras.utils import plot_model\nfrom sklearn.metrics import confusion_matrix\n# -\n\n# # Importing the Data\n\ndata = pd.read_csv(\"EEG_data.csv\", sep=',')\n\ndata.info( )\n\n# +\n#Structure data in pandas dataframe with 64 bits floats\ndf = pd.DataFrame(data, dtype='f8') \n\n#Assign column labels\ndf.columns = [\n 'Subject', \n 'Video', \n 'Attention', \n 'Meditation', \n 'Raw',\n 'Delta', \n 'Theta', \n 'Alpha1', \n 'Alpha2', \n 'Beta1', \n 'Beta2', \n 'Gamma1', \n 'Gamma2', \n 'exp', \n 'obs'\n]\n\n#Check for missing values in df\nmsg_missing = 'Missing values in dataset'\nassert(all(df.isnull() == False)), msg_missing\n\n#Assign student number as major index\nmajor = df['Subject'].values \n\n#Assign video number as minor index\nminor = df['Video'].values\n\n#Create MultiIndex from students and videos\ndf.index = pd.MultiIndex.from_arrays([major, minor])\n\n#Name major and minor index\ndf.index.levels[0].set_names = 'Subject'\ndf.index.levels[1].set_names = 'Video'\n\n#Remove data not part of the analysis\ndel df['Subject']\ndel df['Video']\ndel df['exp']\ndel df['Meditation']\n# del df['Raw']\n\n#Visual check of data\ndf[:10]\n# -\n\n# # Correlation of the variable \n\nimport seaborn as sns\nplt.figure(figsize = (15, 15))\ncorrelation = df.corr()\nsns.heatmap(correlation, vmin = -1.0, square = True, annot = True)\n\ndf['obs'].unique()\n\ndf['obs'].value_counts()\n\nX = np.array(df.drop(['obs'],axis = 1))\ny = np.array(df['obs'])\n\nX.min(), X.max()\n\nfrom sklearn.preprocessing import StandardScaler\n\nX = StandardScaler().fit_transform(X)\n\nX.min(), X.max()\n\ndataset = tf.data.Dataset.from_tensor_slices((X,y))\n\ndataset_size = X.shape[0]\ndataset_size\n\ntrain_size = int(0.70 * dataset_size)\ntrain_size\n\nX_train = dataset.take(train_size)\nX_test = dataset.skip(train_size)\n\nX_train = X_train.shuffle(len(X_train))\n\n# Making batches of 32 \nX_train = X_train.batch(32)\nX_train\n\n# +\n# model building\nimport tensorflow as tf\nfrom tensorflow.keras import Sequential\n\nfrom tensorflow.keras.layers import Embedding\nfrom tensorflow.keras.models import Model\nfrom tensorflow.keras.layers import Input, LSTM, Dense, Attention, Lambda, Dot, Concatenate,Dropout, Activation\n\nfrom tensorflow.keras import backend as K\n\ntf.keras.backend.clear_session()\n\n# +\nfrom tensorflow.keras.layers import Layer\nfrom tensorflow.keras import backend as K\n\n# implementation of Attention Layer\n\n# attention weights are softmax of (v*(w*s