kernel_id
int64
24.2k
23.3M
prompt
stringlengths
8
1.85M
completetion
stringlengths
1
182k
comp_name
stringlengths
5
57
13,482,615
PRE_TRAINING_TIME_START = datetime.now() AUTO = tf.data.experimental.AUTOTUNE<set_options>
cbc=CatBoostClassifier(verbose=0, n_estimators=100) cbc.fit(X_train,y_train) pred_cbc=cbc.predict(X_test)
Titanic - Machine Learning from Disaster
13,482,615
SEED = 42 os.environ['PYTHONHASHSEED']=str(SEED) random.seed(SEED) np.random.seed(SEED) os.environ['TF_DETERMINISTIC_OPS']=str(SEED) tf.random.set_seed(SEED) <set_options>
log=LogisticRegression(max_iter=1000) log.fit(X_train,y_train) pred_log=log.predict(X_test)
Titanic - Machine Learning from Disaster
13,482,615
try: tpu = tf.distribute.cluster_resolver.TPUClusterResolver() print('Running on TPU ', tpu.master()) except ValueError: tpu = None if tpu: tf.config.experimental_connect_to_cluster(tpu) tf.tpu.experimental.initialize_tpu_system(tpu) strategy = tf.distribute.experimental.TPUStrategy(tpu) else: strategy = tf.distribute.get_strategy() print("REPLICAS: ", strategy.num_replicas_in_sync )<define_variables>
print('XGBoost:', round(xgbclassifier.score(X_train, y_train)* 100, 2), '%.\t\t\t RandomForest:', round(randomfc.score(X_train, y_train)* 100, 2), '%.') print('LightGBM:', round(lightgb.score(X_train, y_train)* 100, 2), '%.\t\t\t AdaBoost:', round(ada.score(X_train, y_train)* 100, 2), '%.') print('CatBoost:', round(cbc.score(X_train, y_train)* 100, 2), '%.\t\t\t LogisticRegression:', round(log.score(X_train, y_train)* 100, 2), '%.' )
Titanic - Machine Learning from Disaster
13,482,615
IMAGE_SIZE =(512, 512) GCS_TRAIN_PATHS = [ KaggleDatasets().get_gcs_path('tfrecords'), KaggleDatasets().get_gcs_path('tfrecords-2') ] TRAINING_FILENAMES = [] for i in GCS_TRAIN_PATHS: TRAINING_FILENAMES.append(tf.io.gfile.glob(i + '/*.tfrecords')) TRAINING_FILENAMES = list(itertools.chain.from_iterable(TRAINING_FILENAMES)) GCS_TEST_PATH = KaggleDatasets().get_gcs_path('tfrecords-3') TEST_FILENAMES = tf.io.gfile.glob(GCS_TEST_PATH + '/*.tfrecords') print(len(TRAINING_FILENAMES)) print(len(TEST_FILENAMES))<define_variables>
print('XGBoost:', round(accuracy_score(y_test,pred_xgb)* 100, 2), '%.\t\t\t RandomForest:', round(accuracy_score(y_test,pred_rf)* 100, 2), '%.') print('LightGBM:', round(accuracy_score(y_test,pred_lgb)* 100, 2), '%.\t\t\t AdaBoost:', round(accuracy_score(y_test,pred_ada)* 100, 2), '%.') print('CatBoost:', round(accuracy_score(y_test,pred_cbc)* 100, 2), '%.\t\t\t LogisticRegression:', round(accuracy_score(y_test,pred_log)* 100, 2), '%.' )
Titanic - Machine Learning from Disaster
13,482,615
EPOCHS = 12 DO_AUG = True BATCH_SIZE = 256 current_epoch = 0 chance = 0 NUM_TRAINING_IMAGES = 105390 NUM_TEST_IMAGES = 12186 STEPS_PER_EPOCH = NUM_TRAINING_IMAGES // BATCH_SIZE<define_variables>
X_train=train.drop('Survived',axis=1) X_test=test.copy() y_train = train['Survived'] model_params = { 'n_estimators': randint(4,200), 'max_features': truncnorm(a=0, b=1, loc=0.25, scale=0.1), 'min_samples_split': uniform(0.01, 0.199) } rf_model = RandomForestClassifier() clf = RandomizedSearchCV(rf_model, model_params, n_iter=100, cv=5, random_state=1) model = clf.fit(X_train, y_train) pprint(model.best_estimator_.get_params() )
Titanic - Machine Learning from Disaster
13,482,615
CLASSES = [str(c ).zfill(2)for c in range(0, 42)]<categorify>
testData=test.copy() predictions_rf=model.predict(testData )
Titanic - Machine Learning from Disaster
13,482,615
def decode_image(image_data): image = tf.image.decode_jpeg(image_data, channels=3) image = tf.cast(image, tf.float32)/ 255.0 image = tf.reshape(image, [*IMAGE_SIZE, 3]) return image def read_labeled_tfrecord(example): LABELED_TFREC_FORMAT = { "image": tf.io.FixedLenFeature([], tf.string), "words": tf.io.FixedLenFeature([6633], tf.float32), "label": tf.io.FixedLenFeature([], tf.int64), } example = tf.io.parse_single_example(example, LABELED_TFREC_FORMAT) image = decode_image(example['image']) words = example['words'] label = tf.cast(example['label'], tf.int32) return(( image, words), label) def read_unlabeled_tfrecord(example): UNLABELED_TFREC_FORMAT = { "image": tf.io.FixedLenFeature([], tf.string), "words": tf.io.FixedLenFeature([6633], tf.float32), "filename": tf.io.FixedLenFeature([], tf.string), } example = tf.io.parse_single_example(example, UNLABELED_TFREC_FORMAT) image = decode_image(example['image']) words = example['words'] filename = example['filename'] return(( image, words), filename) def load_dataset(filenames, labeled=True, ordered=False): ignore_order = tf.data.Options() if not ordered: ignore_order.experimental_deterministic = False dataset = tf.data.TFRecordDataset(filenames, num_parallel_reads=AUTO) dataset = dataset.with_options(ignore_order) dataset = dataset.map(read_labeled_tfrecord if labeled else read_unlabeled_tfrecord, num_parallel_calls=AUTO) return dataset def get_training_dataset(do_aug=True): dataset = load_dataset(TRAINING_FILENAMES, labeled=True) if do_aug: dataset = dataset.map(image_augmentation, num_parallel_calls=AUTO) dataset = dataset.repeat() dataset = dataset.shuffle(2048) dataset = dataset.batch(BATCH_SIZE) dataset = dataset.prefetch(AUTO) return dataset def get_test_dataset(ordered=False, tta=None): dataset = load_dataset(TEST_FILENAMES, labeled=False, ordered=ordered) if tta == 0: dataset = dataset.map(image_augmentation_tta_0, num_parallel_calls=AUTO) elif tta == 1: dataset = dataset.map(image_augmentation_tta_1, num_parallel_calls=AUTO) elif tta == 2: dataset = dataset.map(image_augmentation_tta_2, num_parallel_calls=AUTO) dataset = dataset.batch(BATCH_SIZE) dataset = dataset.prefetch(AUTO) return dataset<normalization>
subm=pd.read_csv('.. /input/titanic/gender_submission.csv') Survived=pd.Series(predictions_rf) subm.drop(columns=['Survived'],inplace=True) subm['Survived']=pd.Series(predictions_rf) subm.to_csv(r'RandomForestSubmission.csv', index = False) print("Submitted" )
Titanic - Machine Learning from Disaster
13,181,043
@tf.function def get_mat(rotation, shear, height_zoom, width_zoom, height_shift, width_shift): rotation = math.pi * rotation / 180. shear = math.pi * shear / 180. c1 = tf.math.cos(rotation) s1 = tf.math.sin(rotation) one = tf.constant([1],dtype='float32') zero = tf.constant([0],dtype='float32') rotation_matrix = tf.reshape(tf.concat([c1,s1,zero, -s1,c1,zero, zero,zero,one],axis=0),[3,3]) c2 = tf.math.cos(shear) s2 = tf.math.sin(shear) shear_matrix = tf.reshape(tf.concat([one,s2,zero, zero,c2,zero, zero,zero,one],axis=0),[3,3]) zoom_matrix = tf.reshape(tf.concat([one/height_zoom,zero,zero, zero,one/width_zoom,zero, zero,zero,one],axis=0),[3,3]) shift_matrix = tf.reshape(tf.concat([one,zero,height_shift, zero,one,width_shift, zero,zero,one],axis=0),[3,3]) return K.dot(K.dot(rotation_matrix, shear_matrix), K.dot(zoom_matrix, shift_matrix)) @tf.function def transform(image): DIM = IMAGE_SIZE[0] XDIM = DIM%2 if tf.random.uniform(shape=[], minval=0, maxval=2, dtype=tf.int32, seed=SEED)== 0: rot = 17.* tf.random.normal([1],dtype='float32') else: rot = tf.constant([0],dtype='float32') if tf.random.uniform(shape=[], minval=0, maxval=2, dtype=tf.int32, seed=SEED)== 0: shr = 5.5 * tf.random.normal([1],dtype='float32') else: shr = tf.constant([0],dtype='float32') if tf.random.uniform(shape=[], minval=0, maxval=3, dtype=tf.int32, seed=SEED)== 0: h_zoom = tf.random.normal([1],dtype='float32')/8.5 if h_zoom > 0: h_zoom = 1.0 + h_zoom * -1 else: h_zoom = 1.0 + h_zoom else: h_zoom = tf.constant([1],dtype='float32') if tf.random.uniform(shape=[], minval=0, maxval=3, dtype=tf.int32, seed=SEED)== 0: w_zoom = tf.random.normal([1],dtype='float32')/8.5 if w_zoom > 0: w_zoom = 1.0 + w_zoom * -1 else: w_zoom = 1.0 + w_zoom else: w_zoom = tf.constant([1],dtype='float32') if tf.random.uniform(shape=[], minval=0, maxval=3, dtype=tf.int32, seed=SEED)== 0: h_shift = 18.* tf.random.normal([1],dtype='float32') else: h_shift = tf.constant([0],dtype='float32') if tf.random.uniform(shape=[], minval=0, maxval=3, dtype=tf.int32, seed=SEED)== 0: w_shift = 18.* tf.random.normal([1],dtype='float32') else: w_shift = tf.constant([0],dtype='float32') m = get_mat(rot,shr,h_zoom,w_zoom,h_shift,w_shift) x = tf.repeat(tf.range(DIM//2,-DIM//2,-1), DIM) y = tf.tile(tf.range(-DIM//2,DIM//2),[DIM]) z = tf.ones([DIM*DIM],dtype='int32') idx = tf.stack([x,y,z]) idx2 = K.dot(m,tf.cast(idx,dtype='float32')) idx2 = K.cast(idx2,dtype='int32') idx2 = K.clip(idx2,-DIM//2+XDIM+1,DIM//2) idx3 = tf.stack([DIM//2-idx2[0,], DIM//2-1+idx2[1,]]) d = tf.gather_nd(image,tf.transpose(idx3)) return tf.reshape(d,[DIM,DIM,3] )<prepare_x_and_y>
train_data = pd.read_csv("/kaggle/input/titanic/train.csv") test_data = pd.read_csv("/kaggle/input/titanic/test.csv") train_data
Titanic - Machine Learning from Disaster
13,181,043
@tf.function def transform_grid_mark(image, inv_mat, image_shape): h, w, c = image_shape cx, cy = w//2, h//2 new_xs = tf.repeat(tf.range(-cx, cx, 1), h) new_ys = tf.tile(tf.range(-cy, cy, 1), [w]) new_zs = tf.ones([h*w], dtype=tf.int32) old_coords = tf.matmul(inv_mat, tf.cast(tf.stack([new_xs, new_ys, new_zs]), tf.float32)) old_coords_x, old_coords_y = tf.round(old_coords[0, :] + tf.cast(w, tf.float32)//2.) , tf.round(old_coords[1, :] + tf.cast(h, tf.float32)//2.) old_coords_x = tf.cast(old_coords_x, tf.int32) old_coords_y = tf.cast(old_coords_y, tf.int32) clip_mask_x = tf.logical_or(old_coords_x<0, old_coords_x>w-1) clip_mask_y = tf.logical_or(old_coords_y<0, old_coords_y>h-1) clip_mask = tf.logical_or(clip_mask_x, clip_mask_y) old_coords_x = tf.boolean_mask(old_coords_x, tf.logical_not(clip_mask)) old_coords_y = tf.boolean_mask(old_coords_y, tf.logical_not(clip_mask)) new_coords_x = tf.boolean_mask(new_xs+cx, tf.logical_not(clip_mask)) new_coords_y = tf.boolean_mask(new_ys+cy, tf.logical_not(clip_mask)) old_coords = tf.cast(tf.stack([old_coords_y, old_coords_x]), tf.int32) new_coords = tf.cast(tf.stack([new_coords_y, new_coords_x]), tf.int64) rotated_image_values = tf.gather_nd(image, tf.transpose(old_coords)) rotated_image_channel = list() for i in range(c): vals = rotated_image_values[:,i] sparse_channel = tf.SparseTensor(tf.transpose(new_coords), vals, [h, w]) rotated_image_channel.append(tf.sparse.to_dense(sparse_channel, default_value=0, validate_indices=False)) return tf.transpose(tf.stack(rotated_image_channel), [1,2,0]) @tf.function def random_rotate(image, angle, image_shape): def get_rotation_mat_inv(angle): angle = math.pi * angle / 180 cos_val = tf.math.cos(angle) sin_val = tf.math.sin(angle) one = tf.constant([1], tf.float32) zero = tf.constant([0], tf.float32) rot_mat_inv = tf.concat([cos_val, sin_val, zero, -sin_val, cos_val, zero, zero, zero, one], axis=0) rot_mat_inv = tf.reshape(rot_mat_inv, [3,3]) return rot_mat_inv angle = float(angle)* tf.random.normal([1],dtype='float32') rot_mat_inv = get_rotation_mat_inv(angle) return transform_grid_mark(image, rot_mat_inv, image_shape) @tf.function def grid_mask() : h = tf.constant(IMAGE_SIZE[0], dtype=tf.float32) w = tf.constant(IMAGE_SIZE[1], dtype=tf.float32) image_height, image_width =(h, w) d1 = 112 d2 = 352 rotate_angle = 45 ratio = 0.6 hh = tf.math.ceil(tf.math.sqrt(h*h+w*w)) hh = tf.cast(hh, tf.int32) hh = hh+1 if hh%2==1 else hh d = tf.random.uniform(shape=[], minval=d1, maxval=d2, dtype=tf.int32) l = tf.cast(tf.cast(d,tf.float32)*ratio+0.5, tf.int32) st_h = tf.random.uniform(shape=[], minval=0, maxval=d, dtype=tf.int32) st_w = tf.random.uniform(shape=[], minval=0, maxval=d, dtype=tf.int32) y_ranges = tf.range(-1 * d + st_h, -1 * d + st_h + l) x_ranges = tf.range(-1 * d + st_w, -1 * d + st_w + l) for i in range(0, hh//d+1): s1 = i * d + st_h s2 = i * d + st_w y_ranges = tf.concat([y_ranges, tf.range(s1,s1+l)], axis=0) x_ranges = tf.concat([x_ranges, tf.range(s2,s2+l)], axis=0) x_clip_mask = tf.logical_or(x_ranges <0 , x_ranges > hh-1) y_clip_mask = tf.logical_or(y_ranges <0 , y_ranges > hh-1) clip_mask = tf.logical_or(x_clip_mask, y_clip_mask) x_ranges = tf.boolean_mask(x_ranges, tf.logical_not(clip_mask)) y_ranges = tf.boolean_mask(y_ranges, tf.logical_not(clip_mask)) hh_ranges = tf.tile(tf.range(0,hh), [tf.cast(tf.reduce_sum(tf.ones_like(x_ranges)) , tf.int32)]) x_ranges = tf.repeat(x_ranges, hh) y_ranges = tf.repeat(y_ranges, hh) y_hh_indices = tf.transpose(tf.stack([y_ranges, hh_ranges])) x_hh_indices = tf.transpose(tf.stack([hh_ranges, x_ranges])) y_mask_sparse = tf.SparseTensor(tf.cast(y_hh_indices, tf.int64), tf.zeros_like(y_ranges), [hh, hh]) y_mask = tf.sparse.to_dense(y_mask_sparse, 1, False) x_mask_sparse = tf.SparseTensor(tf.cast(x_hh_indices, tf.int64), tf.zeros_like(x_ranges), [hh, hh]) x_mask = tf.sparse.to_dense(x_mask_sparse, 1, False) mask = tf.expand_dims(tf.clip_by_value(x_mask + y_mask, 0, 1), axis=-1) mask = random_rotate(mask, rotate_angle, [hh, hh, 1]) mask = tf.image.crop_to_bounding_box(mask,(hh-tf.cast(h, tf.int32)) //2,(hh-tf.cast(w, tf.int32)) //2, tf.cast(image_height, tf.int32), tf.cast(image_width, tf.int32)) return mask @tf.function def apply_grid_mask(image): mask = grid_mask() mask = tf.concat([mask, mask, mask], axis=-1) return image * tf.cast(mask, 'float32' )<normalization>
women = train_data.loc[train_data.Sex == 'female']["Survived"] rate_women = sum(women)/len(women) print("% of women who survived:", rate_women*100) men = train_data.loc[train_data.Sex == 'male']["Survived"] rate_men = sum(men)/len(men) print("% of men who survived", rate_men*100) passengers = train_data.Survived.values rate_passengers = sum(passengers)/len(passengers) print(" % of all passengers who survived", rate_passengers*100) print("% of survivors who were women: ", 100*sum(women)/sum(passengers)) print("% of survivors who were men: ", 100*sum(men)/sum(passengers)) print(" % of passengers who were women:", 100*len(train_data.loc[train_data.Sex == 'female'])/len(train_data)) print("% of passengers who were men:",100*len(train_data.loc[train_data.Sex == 'male'])/len(train_data))
Titanic - Machine Learning from Disaster
13,181,043
@tf.function def image_augmentation_tta_0(iw, filename): max_chance = 8 image, words = iw if tf.random.uniform(shape=[], minval=0, maxval=11, dtype=tf.int32, seed=SEED)< max_chance: image = tf.image.flip_left_right(image) if tf.random.uniform(shape=[], minval=0, maxval=11, dtype=tf.int32, seed=SEED)< max_chance: image = tf.image.random_brightness(image, 0.1) if tf.random.uniform(shape=[], minval=0, maxval=11, dtype=tf.int32, seed=SEED)< max_chance: image = tf.image.random_contrast(image, 0.9, 1.1) return(( image, words), filename) @tf.function def image_augmentation_tta_1(iw, filename): max_chance = 8 image, words = iw if tf.random.uniform(shape=[], minval=0, maxval=11, dtype=tf.int32, seed=SEED)< max_chance: image = tf.image.flip_left_right(image) if tf.random.uniform(shape=[], minval=0, maxval=11, dtype=tf.int32, seed=SEED)< max_chance: image = tf.image.random_brightness(image, 0.1) if tf.random.uniform(shape=[], minval=0, maxval=11, dtype=tf.int32, seed=SEED)< max_chance: image = tf.image.random_contrast(image, 0.9, 1.1) if tf.random.uniform(shape=[], minval=0, maxval=11, dtype=tf.int32, seed=SEED)< max_chance: image = tf.image.random_saturation(image, 0.95, 1.05) if tf.random.uniform(shape=[], minval=0, maxval=11, dtype=tf.int32, seed=SEED)< max_chance: image = tf.image.random_hue(image, 0.05) if tf.random.uniform(shape=[], minval=0, maxval=11, dtype=tf.int32, seed=SEED)< max_chance: image = transform(image) return(( image, words), filename) @tf.function def image_augmentation_tta_2(iw, filename): max_chance = 8 image, words = iw if tf.random.uniform(shape=[], minval=0, maxval=11, dtype=tf.int32, seed=SEED)< max_chance: image = tf.image.flip_left_right(image) if tf.random.uniform(shape=[], minval=0, maxval=11, dtype=tf.int32, seed=SEED)< max_chance: image = tf.image.random_brightness(image, 0.1) if tf.random.uniform(shape=[], minval=0, maxval=11, dtype=tf.int32, seed=SEED)< max_chance: image = tf.image.random_contrast(image, 0.9, 1.1) if tf.random.uniform(shape=[], minval=0, maxval=11, dtype=tf.int32, seed=SEED)< max_chance: image = tf.image.random_saturation(image, 0.95, 1.05) if tf.random.uniform(shape=[], minval=0, maxval=11, dtype=tf.int32, seed=SEED)< max_chance: image = tf.image.random_hue(image, 0.05) if tf.random.uniform(shape=[], minval=0, maxval=11, dtype=tf.int32, seed=SEED)< max_chance: image = transform(image) if tf.random.uniform(shape=[], minval=0, maxval=10, dtype=tf.int32, seed=SEED)< max_chance: image = apply_grid_mask(image) return(( image, words), filename )<import_modules>
withsbsp=train_data.loc[train_data.Survived == 0]["SibSp"] print("Average number of siblings/spouses for those who died: {}".format(sum(withsbsp)/len(withsbsp))) survivedsbsp = train_data.loc[train_data.Survived == 1]["SibSp"] print("Average number of siblings/spouses for those who survived: {}".format(sum(survivedsbsp)/len(survivedsbsp))) nmen = train_data.loc[train_data.Survived == 0][train_data.Sex == 'male']["Parch"] ymen = train_data.loc[train_data.Survived == 1][train_data.Sex == 'male']["Parch"] print(" Average number of children/parents per man who did not survive: {}".format(sum(nmen)/len(nmen))) print("Average number of children/parents per man who survived: {}".format(sum(ymen)/len(ymen)))
Titanic - Machine Learning from Disaster
13,181,043
from tensorflow.keras.models import Model, Sequential from tensorflow.keras.layers import Input, Flatten, Dense, Dropout, AveragePooling2D, GlobalAveragePooling2D, SpatialDropout2D, BatchNormalization, Activation, Concatenate<init_hyperparams>
male=train_data.loc[train_data.Sex=='male'] nmen = train_data.loc[train_data.Survived == 0][train_data.Sex == 'male'][train_data.Parch >=1] ymen = train_data.loc[train_data.Survived == 1][train_data.Sex == 'male'][train_data.Parch >=1] mediann=np.median(nmen.Parch) mediany=np.median(ymen.Parch) print(" Average number of children/Fathers per man who did not survive: {} The median is {}" .format(sum(nmen.Parch)/len(nmen),mediann)) print("Average number of children/Fathers per man who survived: {} The median is {}" .format(sum(ymen.Parch)/len(ymen),mediany)) print("% of men with children/parents: ", 100*(len(nmen)+len(ymen)) /len(male))
Titanic - Machine Learning from Disaster
13,181,043
class EpochCallback(tf.keras.callbacks.Callback): def on_epoch_begin(self, epoch, logs=None): global current_epoch global chance current_epoch = epoch if current_epoch < 2: chance = 0 elif current_epoch < 9: chance = current_epoch - 1 else: chance = 8 print(f'Epoch print(datetime.now() )<choose_model_class>
train_data1 = train_data.drop(['Name', 'Ticket','Cabin','Embarked','PassengerId'], axis = 1) train_data1 X_test = test_data.drop(['Name', 'Ticket','Cabin','Embarked','PassengerId'], axis = 1) train_data1 = pd.get_dummies(train_data1) train_data1 = train_data1.drop(['Sex_male'],axis=1) X_test=pd.get_dummies(X_test) X_test= X_test.drop(['Sex_male'],axis=1)
Titanic - Machine Learning from Disaster
13,181,043
es_val_acc = tf.keras.callbacks.EarlyStopping( monitor='val_sparse_categorical_accuracy', min_delta=0.001, patience=5, verbose=1, mode='auto', baseline=None, restore_best_weights=True ) es_val_loss = tf.keras.callbacks.EarlyStopping( monitor='val_loss', min_delta=0.001, patience=5, verbose=1, mode='auto', baseline=None, restore_best_weights=True ) es_acc = tf.keras.callbacks.EarlyStopping( monitor='sparse_categorical_accuracy', min_delta=0.001, patience=5, verbose=1, mode='auto', baseline=None, restore_best_weights=False ) es_loss = tf.keras.callbacks.EarlyStopping( monitor='loss', min_delta=0.001, patience=5, verbose=1, mode='auto', baseline=None, restore_best_weights=False ) epoch_cb = EpochCallback()<choose_model_class>
c1a = int(( train_data1.loc[train_data1.Pclass ==1])['Age'].mean()) c2a = int(( train_data1.loc[train_data1.Pclass ==2])['Age'].mean()) c3a = int(( train_data1.loc[train_data1.Pclass ==3])['Age'].mean()) print(c1a,c2a,c3a) a =train_data1.loc[train_data.Pclass==1].fillna({'Age':c1a}) train_data1.loc[train_data.Pclass==1]=a a2=X_test.loc[X_test.Pclass==1].fillna({'Age':c1a}) X_test.loc[X_test.Pclass==1] =a2 b=train_data1.loc[train_data.Pclass==2].fillna({'Age':c2a}) train_data1.loc[train_data.Pclass==2]=b b2=X_test.loc[X_test.Pclass==2].fillna({'Age':c2a}) X_test.loc[X_test.Pclass==2] =b2 c=train_data1.loc[train_data.Pclass==3].fillna({'Age':c3a}) train_data1.loc[train_data.Pclass==3]=c c2=X_test.loc[X_test.Pclass==3].fillna({'Age':c3a}) X_test.loc[X_test.Pclass==3] =c2 X_test['Fare'] = X_test['Fare'].fillna(20)
Titanic - Machine Learning from Disaster
13,181,043
with strategy.scope() : efn7 = efn.EfficientNetB7(weights='noisy-student', include_top=False, input_shape=(IMAGE_SIZE[0], IMAGE_SIZE[1], 3)) for layer in efn7.layers: layer.trainable = True model_image = Sequential([ efn7, GlobalAveragePooling2D(name='efficientnet-b7_gap'), ], name='b7-image') model_words = Sequential([ Input(( 6633,), name='mlp-words_input'), Dense(331, name='mlp-words_dense_1'), BatchNormalization(name='mlp-words_bn_1'), Activation('relu', name='mlp-words_act_1'), Dense(110, name='mlp-words_dense_2'), BatchNormalization(name='mlp-words_bn_2'), Activation('relu', name='mlp-words_act_2'), ], name='mlp-words') concatenate = Concatenate(name='concatenate' )([model_image.output, model_words.output]) output = Dense(len(CLASSES), activation='softmax', name='output' )(concatenate) model = Model(inputs=[model_image.input, model_words.input], outputs=output) model.compile(optimizer=tfa.optimizers.LAMB(0.01), loss='sparse_categorical_crossentropy', metrics=['sparse_categorical_accuracy']) model.summary() <train_model>
X_train = train_data1.drop(columns=["Survived"]) y_train = train_data1.Survived params = {'n_estimators': [1,200], 'max_features': [1,6], 'max_depth': [1,None], 'min_samples_split': [2,10], 'min_samples_leaf': [1,10], 'criterion':['gini','entropy'] } forest = RandomForestClassifier() forest_cv = RandomizedSearchCV(forest, params, cv=5) forest_cv.fit(X_train,y_train) print("Tuned Decision Tree Parameters: {}".format(forest_cv.best_params_)) print("Best score is {}".format(forest_cv.best_score_))
Titanic - Machine Learning from Disaster
13,181,043
<init_hyperparams><EOS>
y_pred = forest_cv.predict(X_test) output = pd.DataFrame({'PassengerId': test_data.PassengerId, 'Survived': y_pred}) output.to_csv('my_submission.csv', index=False) print("Your submission was successfully saved!" )
Titanic - Machine Learning from Disaster
13,359,181
<SOS> metric: categorizationaccuracy Kaggle data source: titanic-machine-learning-from-disaster<train_model>
%matplotlib inline warnings.filterwarnings('ignore' )
Titanic - Machine Learning from Disaster
13,359,181
model.fit( get_training_dataset(do_aug=DO_AUG), steps_per_epoch=STEPS_PER_EPOCH, epochs=EPOCHS, callbacks=[es_acc, epoch_cb, lr_schedule], verbose=1 )<compute_test_metric>
train = pd.read_csv(".. /input/titanic/train.csv") test = pd.read_csv(".. /input/titanic/test.csv") train.describe(include="all" )
Titanic - Machine Learning from Disaster
13,359,181
h = model.history plt_acc(h) plt_loss(h )<feature_engineering>
ids = test['PassengerId']
Titanic - Machine Learning from Disaster
13,359,181
POST_TRAINING_TIME_START = datetime.now()<predict_on_test>
Cat_Features = ['Sex', 'Pclass', 'SibSp', 'Parch', 'Embarked']
Titanic - Machine Learning from Disaster
13,359,181
def test(tta=None): test_ds = get_test_dataset(ordered=True, tta=tta) print(f'Computing predictions for TTA {tta}...') test_images_ds = test_ds.map(lambda iw, filename: [iw]) model_pred = model.predict(test_images_ds) return model_pred<categorify>
print(pd.isnull(train ).sum() )
Titanic - Machine Learning from Disaster
13,359,181
model_pred = test(tta=None) model_pred_tta_0 = test(tta=0) model_pred_tta_1 = test(tta=1) model_pred_tta_2 = test(tta=2 )<prepare_output>
all_data = pd.concat([train.drop(columns='Survived'), test], ignore_index=True) print(all_data.shape )
Titanic - Machine Learning from Disaster
13,359,181
pred_plain = np.argmax(model_pred, axis=-1) test_ds = get_test_dataset(ordered=True) test_ids_ds = test_ds.map(lambda iw, filename: filename ).unbatch() test_ids = next(iter(test_ids_ds.batch(pred_plain.shape[0])) ).numpy().astype('U') df_submission = pd.DataFrame({'filename': test_ids, 'category': pred_plain}) df_submission = df_submission.drop_duplicates() df_submission['category'] = df_submission['category'].apply(lambda c: str(c ).zfill(2)) df_submission<save_to_csv>
for feature in ['PassengerId','Cabin', 'Ticket']: all_data.drop(feature, axis = 1, inplace=True) all_data.head()
Titanic - Machine Learning from Disaster
13,359,181
df_submission.to_csv('submission.csv', index=False) !head submission.csv<feature_engineering>
all_data['Embarked'].fillna(all_data['Embarked'].mode() [0], inplace = True) all_data['Fare'].fillna(all_data['Fare'].median() , inplace = True)
Titanic - Machine Learning from Disaster
13,359,181
pred_tta_0 = np.mean(np.array([model_pred, model_pred_tta_0]), axis=0) pred_tta_0 = np.argmax(pred_tta_0, axis=-1) test_ds = get_test_dataset(ordered=True) test_ids_ds = test_ds.map(lambda iw, filename: filename ).unbatch() test_ids = next(iter(test_ids_ds.batch(pred_tta_0.shape[0])) ).numpy().astype('U') df_submission_tta_0 = pd.DataFrame({'filename': test_ids, 'category': pred_tta_0}) df_submission_tta_0 = df_submission_tta_0.drop_duplicates() df_submission_tta_0['category'] = df_submission_tta_0['category'].apply(lambda c: str(c ).zfill(2)) df_submission_tta_0<save_to_csv>
all_data['Title'] = all_data.Name.str.extract('([A-Za-z]+)\.', expand=False) all_data['Title'].value_counts()
Titanic - Machine Learning from Disaster
13,359,181
df_submission_tta_0.to_csv('submission_tta_0.csv', index=False) !head submission_tta_0.csv<feature_engineering>
frequent_titles = all_data['Title'].value_counts() [:5].index.tolist() frequent_titles
Titanic - Machine Learning from Disaster
13,359,181
pred_tta_all = np.mean(np.array([model_pred, model_pred_tta_0, model_pred_tta_1, model_pred_tta_2]), axis=0) pred_tta_all = np.argmax(pred_tta_all, axis=-1) test_ds = get_test_dataset(ordered=True) test_ids_ds = test_ds.map(lambda iw, filename: filename ).unbatch() test_ids = next(iter(test_ids_ds.batch(pred_tta_all.shape[0])) ).numpy().astype('U') df_submission_tta_all = pd.DataFrame({'filename': test_ids, 'category': pred_tta_all}) df_submission_tta_all = df_submission.drop_duplicates() df_submission_tta_all['category'] = df_submission_tta_all['category'].apply(lambda c: str(c ).zfill(2)) df_submission_tta_all<save_to_csv>
all_data['Title'] = all_data['Title'].apply(lambda x: x if x in frequent_titles else 'Other')
Titanic - Machine Learning from Disaster
13,359,181
df_submission_tta_all.to_csv('submission_tta_all.csv', index=False) !head submission_tta_all.csv<load_pretrained>
median_ages = {} for title in frequent_titles: median_ages[title] = all_data.loc[all_data['Title'] == title]['Age'].median() median_ages['Other'] = all_data['Age'].median() all_data.loc[all_data['Age'].isnull() , 'Age'] = all_data[all_data['Age'].isnull() ]['Title'].map(median_ages )
Titanic - Machine Learning from Disaster
13,359,181
model.save('model.h5') model.save_weights('model_weights.h5' )<train_model>
num_bins = 5 feature = 'Fare' labels = [feature + '_' + str(i + 1)for i in range(num_bins)] all_data['FareBin'] = pd.qcut(all_data['Fare'], num_bins, labels )
Titanic - Machine Learning from Disaster
13,359,181
print(f'Post training time : {(datetime.now() - POST_TRAINING_TIME_START ).total_seconds() } seconds' )<define_variables>
used_features = ['Age', 'Fare','Name'] for feature in used_features: all_data.drop(feature, axis = 1, inplace=True )
Titanic - Machine Learning from Disaster
13,359,181
numeric_features_train = list(train.select_dtypes(exclude = ['object', 'category'] ).columns) numeric_features_test = list(test.select_dtypes(exclude = ['object', 'category'] ).columns) print("numeric_features_train = ", numeric_features_train) print("Len = ", len(numeric_features_train)) print(" numeric_features_test = ", numeric_features_test) print("Len = ", len(numeric_features_test))<count_missing_values>
all_data = pd.get_dummies(all_data) all_data.head()
Titanic - Machine Learning from Disaster
13,359,181
print(train.isnull().any()) print(test.isnull().any() )<drop_column>
target = train['Survived'] train = all_data.iloc[: len(train)] test = all_data.iloc[len(train):] train.shape
Titanic - Machine Learning from Disaster
13,359,181
SP = train.pop('SalePrice' )<split>
MLA = [ ensemble.AdaBoostClassifier() , ensemble.BaggingClassifier() , ensemble.ExtraTreesClassifier() , ensemble.GradientBoostingClassifier() , ensemble.RandomForestClassifier() , gaussian_process.GaussianProcessClassifier() , linear_model.LogisticRegressionCV() , linear_model.PassiveAggressiveClassifier() , linear_model.RidgeClassifierCV() , linear_model.SGDClassifier() , linear_model.Perceptron() , naive_bayes.BernoulliNB() , naive_bayes.GaussianNB() , neighbors.KNeighborsClassifier() , svm.SVC(probability=True), svm.NuSVC(probability=True), svm.LinearSVC() , tree.DecisionTreeClassifier() , tree.ExtraTreeClassifier() , discriminant_analysis.LinearDiscriminantAnalysis() , discriminant_analysis.QuadraticDiscriminantAnalysis() , XGBClassifier() ] cv_split = model_selection.ShuffleSplit(n_splits = 10, test_size =.3, random_state = 0) MLA_columns = ['MLA Name', 'MLA Train Accuracy Mean', 'MLA Validation Accuracy Mean', 'MLA Validation Accuracy STD'] MLA_compare = pd.DataFrame(columns = MLA_columns) estimators = [] for row_index, alg in enumerate(MLA): alg_name = alg.__class__.__name__ MLA_compare.loc[row_index, 'MLA Name'] = alg_name print(alg.__class__.__name__) cv_results = model_selection.cross_validate(alg, train, target, cv=cv_split, return_estimator=True, return_train_score=True) estimators += list(cv_results['estimator']) MLA_compare.loc[row_index, 'MLA Train Accuracy Mean'] = cv_results['train_score'].mean() MLA_compare.loc[row_index, 'MLA Validation Accuracy Mean'] = cv_results['test_score'].mean() MLA_compare.loc[row_index, 'MLA Validation Accuracy STD'] = cv_results['test_score'].std() MLA_compare.sort_values(by = ['MLA Validation Accuracy Mean'], ascending = False, inplace = True) MLA_compare
Titanic - Machine Learning from Disaster
13,359,181
def getTheta(r): X_train, X_test , y_train, y_test = train_test_split(train_X, SP, test_size = 0.2, random_state = r) theta = np.dot(np.linalg.pinv(X_train), y_train) print("Training Error =",rmse(predictTarget(X_train, theta), y_train)) print("Validation Error = ", rmse(predictTarget(X_test, theta), y_test)) return theta<predict_on_test>
def hard_voting(estimators, X): Y = np.zeros([X.shape[0], len(estimators)], dtype=int) for i, estimator in enumerate(estimators): Y[:, i] = estimator.predict(X) y = np.zeros(X.shape[0], dtype=int) for i in range(X.shape[0]): y[i] = np.argmax(np.bincount(Y[i,:])) return y y_predict = hard_voting(estimators, train) accuracy_score(target, y_predict)
Titanic - Machine Learning from Disaster
13,359,181
<predict_on_test><EOS>
predictions = hard_voting(estimators, test) output = pd.DataFrame({ 'PassengerId' : ids, 'Survived': predictions }) output.to_csv('submission_voting_cross_validation.csv', index=False )
Titanic - Machine Learning from Disaster
13,410,275
<SOS> metric: categorizationaccuracy Kaggle data source: titanic-machine-learning-from-disaster<predict_on_test>
for dirname, _, filenames in os.walk('/kaggle/input'): for filename in filenames: print(os.path.join(dirname, filename))
Titanic - Machine Learning from Disaster
13,410,275
prediction = np.zeros(( test_X.shape[0], 1)) rList = [42, 3, 468, 12, 120, 130, 234, 1000, 100, 200, 1200, 5, 45, 49, 300, 320,980, 81, 2100, 1420, 768] for r in rList: y1 = predictTarget(test_X, getTheta(r)) print(prediction.shape) prediction = np.hstack(( prediction,y1.reshape(y1.shape[0], 1)) )<prepare_output>
ds_train = pd.read_csv('/kaggle/input/titanic/train.csv') ds_output = pd.read_csv('/kaggle/input/titanic/test.csv' )
Titanic - Machine Learning from Disaster
13,410,275
prediction = np.sum(prediction, axis = 1)/len(rList )<save_to_csv>
X_train = ds_train.drop(['PassengerId','Name','Ticket'],axis=1) X_output = ds_output.drop(['PassengerId','Name','Ticket'],axis=1 )
Titanic - Machine Learning from Disaster
13,410,275
sub = pd.DataFrame() sub['Id'] = test['Id'] sub['SalePrice'] = prediction sub.to_csv("prediction.csv", index = False )<set_options>
X_train['Cabin'].isna().sum() / X_train.shape[0]
Titanic - Machine Learning from Disaster
13,410,275
%matplotlib inline sns.set()<load_from_csv>
X_train = X_train.drop(['Cabin'],axis=1) X_output = X_output.drop(['Cabin'],axis=1 )
Titanic - Machine Learning from Disaster
13,410,275
sub = pd.read_csv('.. /input/sample_submission.csv' )<load_from_disk>
print(f"Count : {X_train['Embarked'].isna().sum() } / {X_train.shape[0]} = {X_train['Embarked'].isna().sum() /X_train.shape[0]}") print(f"Most common letter : {np.squeeze(X_train['Embarked'].mode())}" )
Titanic - Machine Learning from Disaster
13,410,275
with open('.. /input/train_annotations.json/train_annotations.json','r')as anno_train: train = json.load(anno_train) with open('.. /input/test_annotations.json/test_annotations.json','r')as anno_test: test = json.load(anno_test )<create_dataframe>
letter_dictionnary = X_train['Embarked'].astype('category' ).cat.categories
Titanic - Machine Learning from Disaster
13,410,275
test_df = pd.DataFrame() test_df = test_df.append(test['images'], ignore_index=True )<create_dataframe>
X_train['Sex']=X_train['Sex'].astype('category' ).cat.codes X_output["Sex"]=X_output["Sex"].astype('category' ).cat.codes
Titanic - Machine Learning from Disaster
13,410,275
train_df = pd.DataFrame() train_df = train_df.append(train['images'], ignore_index=True) train_df_anno = pd.DataFrame() train_df_anno = train_df_anno.append(train['annotations'], ignore_index=True) train_df['category_id'] = train_df_anno['category_id'] del train_df_anno gc.collect()<define_variables>
X_train['Embarked']=X_train['Embarked'].astype('category' ).cat.codes X_output['Embarked']=X_output["Embarked"].astype('category' ).cat.codes
Titanic - Machine Learning from Disaster
13,410,275
category_numbers = [0]*(np.max(train_df['category_id'])+1 )<feature_engineering>
X_train = pd.concat([X_train, pd.get_dummies(X_train.Pclass)], axis=1) X_output = pd.concat([X_output, pd.get_dummies(X_output.Pclass)], axis=1) X_train.rename(columns={1: "Pclass1", 2: "Pclass2",3:"Pclass3"}, inplace=True) X_output.rename(columns={1: "Pclass1", 2: "Pclass2",3:"Pclass3"}, inplace=True) X_train.drop(['Pclass'],inplace=True,axis=1) X_output.drop(['Pclass'],inplace=True,axis=1 )
Titanic - Machine Learning from Disaster
13,410,275
for i in train_df['category_id']: category_numbers[i]+=1<define_variables>
X_train = pd.concat([X_train, pd.get_dummies(X_train.Embarked)], axis=1) X_output = pd.concat([X_output, pd.get_dummies(X_output.Embarked)], axis=1) X_train.rename(columns={0: f"Embarked_{letter_dictionnary[0]}", 1: f"Embarked_{letter_dictionnary[1]}",2: f"Embarked_{letter_dictionnary[2]}"}, inplace=True) X_output.rename(columns={0: f"Embarked_{letter_dictionnary[0]}", 1: f"Embarked_{letter_dictionnary[1]}",2: f"Embarked_{letter_dictionnary[2]}"}, inplace=True) X_train.drop(['Embarked'],inplace=True,axis=1) X_output.drop(['Embarked'],inplace=True,axis=1 )
Titanic - Machine Learning from Disaster
13,410,275
few_shot = 0 for i in category_numbers[1:]: if i <= 200: few_shot+=1<feature_engineering>
def normalization(column): return(column-column.mean())/column.std()
Titanic - Machine Learning from Disaster
13,410,275
few_shot/len(category_numbers )<load_from_csv>
X_train['Age'],X_train['Fare'] = normalization(X_train['Age']),normalization(X_train['Fare'] )
Titanic - Machine Learning from Disaster
13,410,275
df = pd.read_csv('.. /input/train.csv') test = pd.read_csv('.. /input/test.csv' )<concatenate>
X_output['Age'],X_output['Fare'] = normalization(X_output['Age']),normalization(X_output['Fare'] )
Titanic - Machine Learning from Disaster
13,410,275
df = df.append(test )<data_type_conversions>
y_train = X_train['Survived'] X_train = X_train.drop(['Survived'], axis=1 )
Titanic - Machine Learning from Disaster
13,410,275
df['area'] = df['area'].str.replace(',','') df['densidade_dem'] = df['densidade_dem'].str.replace(',','') df['area'] = df['area'].astype(float) df['densidade_dem'] = df['densidade_dem'].astype(float )<feature_engineering>
X_train.reset_index(drop=True, inplace=True) y_train.reset_index(drop=True, inplace=True) X_output.reset_index(drop=True, inplace=True )
Titanic - Machine Learning from Disaster
13,410,275
df['populacao'] = df['populacao'].str.replace(',','') df['populacao'] = df['populacao'].str.replace('(2)','') df['populacao'] = df['populacao'].str.replace('(','') df['populacao'] = df['populacao'].str.replace(')','') df['populacao'] = df['populacao'].str.replace('.','') df['populacao'] = df['populacao'].astype(int )<feature_engineering>
X_train,X_test,y_train,y_test = train_test_split(X_train,y_train )
Titanic - Machine Learning from Disaster
13,410,275
df['comissionados_por_servidor'] = df['comissionados_por_servidor'].str.replace(' df['comissionados_por_servidor'] = df['comissionados_por_servidor'].str.replace('!','') df['comissionados_por_servidor'] = df['comissionados_por_servidor'].str.replace('%','') df['comissionados_por_servidor'] = df['comissionados_por_servidor'].astype(float) df['comissionados_por_servidor'] = df['comissionados_por_servidor']/100<categorify>
hyperparameters = {'C': np.arange(0.5,4,0.2), 'kernel':['linear', 'rbf', 'sigmoid'], 'gamma':np.linspace(10**-3,1,50)} grid = GridSearchCV(svm.SVC() , hyperparameters, cv=5) grid.fit(X_train,y_train) C, kernel, gamma = grid.best_params_['C'],grid.best_params_['kernel'],grid.best_params_['gamma'] print(f"Best score : {grid.best_score_} for hyperparameters : {grid.best_params_}" )
Titanic - Machine Learning from Disaster
13,410,275
df_porte = pd.get_dummies(df['porte']) df_regiao = pd.get_dummies(df['regiao'] )<concatenate>
KM = KMeans(n_clusters=30) KM.fit(X_train,y_train) print(f"Best score : {completeness_score(KM.predict(X_test),y_test)} for hyperparameters : n_clusters = 15" )
Titanic - Machine Learning from Disaster
13,410,275
df = pd.concat([df, df_porte], axis=1) df = pd.concat([df, df_regiao], axis=1 )<import_modules>
hyperparameters = {'n_neighbors': np.arange(1,50), 'metric': ['euclidean','manhattan']} grid = GridSearchCV(KNeighborsClassifier() , hyperparameters, cv=5) grid.fit(X_train,y_train) n_neighbors,metric = grid.best_params_['n_neighbors'],grid.best_params_['metric'] print(f"Best score : {grid.best_score_} for hyperparameters : {grid.best_params_}" )
Titanic - Machine Learning from Disaster
13,410,275
import matplotlib.pyplot as plt import seaborn as sns<split>
hyperparameters = {'n_estimators': [1500,2000,2500,3000,4000],'max_features':['auto','sqrt','log2'], 'max_depth':[10,20,30]} grid = GridSearchCV(RandomForestClassifier() , hyperparameters, cv=5) grid.fit(X_train,y_train) n_estimators,max_features,max_depth = grid.best_params_['n_estimators'],grid.best_params_['max_features'],grid.best_params_['max_depth'] print(f"Best score : {grid.best_score_} for hyperparameters : {grid.best_params_}" )
Titanic - Machine Learning from Disaster
13,410,275
test = df[df['nota_mat'].isnull() ] <filter>
hyperparameters = {'n_estimators': [500,700,800,900,1000],'max_features':['auto','sqrt','log2'], 'max_depth':[10,20,30]} grid = GridSearchCV(ExtraTreesClassifier() , hyperparameters, cv=5) grid.fit(X_train,y_train) n_e,max_f,max_d= grid.best_params_['n_estimators'],grid.best_params_['max_features'],grid.best_params_['max_depth'] print(f"Best score : {grid.best_score_} for hyperparameters : {grid.best_params_}" )
Titanic - Machine Learning from Disaster
13,410,275
df = df[~df['nota_mat'].isnull() ]<split>
best_mean_score = 0
Titanic - Machine Learning from Disaster
13,410,275
train, valid = train_test_split(df, random_state=42 )<define_variables>
sv = svm.SVC(C=C, kernel=kernel, gamma=gamma) sv.fit(X_train,y_train) sv_mean_score, sv_std_score = cross_val_score(sv,X_test, y_test, cv = 5 ).mean() ,cross_val_score(sv,X_test, y_test, cv = 5 ).std() print(sv_mean_score, sv_std_score) if sv_mean_score>best_mean_score: best_mean_score = sv_mean_score model = "sv"
Titanic - Machine Learning from Disaster
13,410,275
removed_cols = ['municipio', 'estado', 'codigo_mun', 'porte', 'regiao', 'nota_mat','capital'] feats = [c for c in df.columns if c not in removed_cols]<import_modules>
knn = KNeighborsClassifier(n_neighbors, metric=metric) knn.fit(X_train,y_train) knn_mean_score, knn_std_score = cross_val_score(knn,X_test, y_test, cv = 5 ).mean() ,cross_val_score(knn,X_test, y_test, cv = 5 ).std() print(knn_mean_score, knn_std_score) if knn_mean_score>best_mean_score: best_mean_score = knn_mean_score model = "knn"
Titanic - Machine Learning from Disaster
13,410,275
from sklearn.ensemble import RandomForestClassifier, ExtraTreesClassifier from sklearn.ensemble import AdaBoostClassifier, GradientBoostingClassifier from sklearn.tree import DecisionTreeClassifier from sklearn.neighbors import KNeighborsClassifier<choose_model_class>
rdf = RandomForestClassifier(n_estimators=n_estimators, max_depth=max_depth, max_features=max_features) rdf.fit(X_train,y_train) rdf_mean_score, rdf_std_score = cross_val_score(rdf, X_test, y_test, cv = 5 ).mean() ,cross_val_score(rdf, X_test, y_test, cv = 5 ).std() print(rdf_mean_score, rdf_std_score) if rdf_mean_score>best_mean_score: best_mean_score = rdf_mean_score model = "rdf"
Titanic - Machine Learning from Disaster
13,410,275
models = {'RandomForest': RandomForestClassifier(random_state=42, min_samples_leaf=8, min_samples_split=3, n_jobs=-1, n_estimators=75), 'ExtraTrees': ExtraTreesClassifier(random_state=42, n_jobs=-1), 'GradientBoosting': GradientBoostingClassifier(random_state=42), 'DecisionTree': DecisionTreeClassifier(random_state=42), 'AdaBoost': AdaBoostClassifier(random_state=42), 'KNM 11': KNeighborsClassifier(n_neighbors=11, n_jobs=-1),}<import_modules>
clf = ExtraTreesClassifier(n_estimators=n_e, max_depth=max_d,max_features=max_f) clf.fit(X_train,y_train) clf_mean_score, clf_std_score = cross_val_score(clf, X_test, y_test, cv = 5 ).mean() ,cross_val_score(clf, X_test, y_test, cv = 5 ).std() print(clf_mean_score, clf_std_score) if clf_mean_score>best_mean_score: best_mean_score = clf_mean_score model = "clf"
Titanic - Machine Learning from Disaster
13,410,275
from sklearn.metrics import accuracy_score<predict_on_test>
if model=="svm": y_output = sv.predict(X_output) elif model=="knn": y_output = knn.predict(X_output) elif model=="rdf": y_output = rdf.predict(X_output) else: y_output = clf.predict(X_output) print("Chosen model :", model )
Titanic - Machine Learning from Disaster
13,410,275
def run_model(models, train, valid, feats, y_name): model.fit(train[feats], train[y_name]) preds = model.predict(valid[feats]) return accuracy_score(valid[y_name], preds )<find_best_params>
submission = ds_output["PassengerId"]
Titanic - Machine Learning from Disaster
13,410,275
scores = [] for name, model in models.items() : score = run_model(model, train.fillna(-1), valid.fillna(-1), feats, 'nota_mat') scores.append(score) print(name, ':', score )<concatenate>
submission = pd.DataFrame(submission )
Titanic - Machine Learning from Disaster
13,410,275
train = train.append(valid )<choose_model_class>
submission['Survived']=y_output
Titanic - Machine Learning from Disaster
13,410,275
rf = RandomForestClassifier(random_state=42, min_samples_leaf=8, min_samples_split=3, n_jobs=-1, n_estimators=200 )<train_model>
submission.set_index('PassengerId',drop=True, inplace=True )
Titanic - Machine Learning from Disaster
13,410,275
rf.fit(train[feats].fillna(-1), train['nota_mat'] )<predict_on_test>
submission.to_csv(r'/kaggle/working/submission.csv' )
Titanic - Machine Learning from Disaster
13,055,333
test['nota_mat'] = rf.predict(test[feats].fillna(-1))<save_to_csv>
train_data = pd.read_csv(".. /input/titanic/train.csv") train_data.head()
Titanic - Machine Learning from Disaster
13,055,333
test[['codigo_mun','nota_mat']].to_csv('randomforest.csv', index=False )<load_from_csv>
test_data = pd.read_csv(".. /input/titanic/test.csv") test_data.head()
Titanic - Machine Learning from Disaster
13,055,333
df = pd.read_csv('.. /input/train.csv') df.shape<load_from_csv>
train_data["relatives"] = train_data["SibSp"] + train_data["Parch"] test_data["relatives"] = test_data["SibSp"] + test_data["Parch"]
Titanic - Machine Learning from Disaster
13,055,333
df_test = pd.read_csv('.. /input/test.csv') df_test.shape<concatenate>
train_data["Sex_cat"] = train_data["Sex"].astype("category" )
Titanic - Machine Learning from Disaster
13,055,333
df = df.append(df_test, ignore_index=True, sort=False) df.shape<data_type_conversions>
test_data["Sex_cat"] = test_data["Sex"].astype("category" )
Titanic - Machine Learning from Disaster
13,055,333
df['populacao'] = df['populacao'].str.replace('.', '') df['populacao'] = df['populacao'].str.replace('\(\)', '') df['populacao'] = df['populacao'].str.replace('\(1\)', '') df['populacao'] = df['populacao'].str.replace('\(2\)', '') df['populacao'] = df['populacao'].str.replace(',', '' ).astype('float') df['densidade_dem'] = df['densidade_dem'].str.replace(',', '' ).astype('float') df['area'] = df['area'].str.replace(',', '' ).astype('float') df['comissionados_por_servidor'] = df['comissionados_por_servidor'].str.replace(' df['comissionados_por_servidor'] = df['comissionados_por_servidor'].str.replace('%', '' ).astype('float' )<data_type_conversions>
train_data["Embarked_filled"] = train_data["Embarked_filled"].astype("category") test_data["Embarked_filled"] = test_data["Embarked_filled"].astype("category" )
Titanic - Machine Learning from Disaster
13,055,333
df['in_na_densidade_dem'] = df['densidade_dem'].isna().astype('int') df['densidade_dem'].fillna(df['densidade_dem'].mean() , inplace=True) df['in_na_participacao_transf_receita'] = df['participacao_transf_receita'].isna().astype('int') df['participacao_transf_receita'].fillna(-1, inplace=True) df['in_na_servidores'] = df['servidores'].isna().astype('int') df['servidores'].fillna(-1, inplace=True) df['in_na_perc_pop_econ_ativa'] = df['perc_pop_econ_ativa'].isna().astype('int') df['perc_pop_econ_ativa'].fillna(-1, inplace=True) df['in_na_gasto_pc_saude'] = df['gasto_pc_saude'].isna().astype('int') df['gasto_pc_saude'].fillna(-1, inplace=True) df['in_na_hab_p_medico'] = df['hab_p_medico'].isna().astype('int') df['hab_p_medico'].fillna(-1, inplace=True) df['in_na_exp_vida'] = df['exp_vida'].isna().astype('int') df['exp_vida'].fillna(df['exp_vida'].mean() , inplace=True) df['in_na_gasto_pc_educacao'] = df['gasto_pc_educacao'].isna().astype('int') df['gasto_pc_educacao'].fillna(-1, inplace=True) df['in_na_exp_anos_estudo'] = df['exp_anos_estudo'].isna().astype('int') df['exp_anos_estudo'].fillna(-1, inplace=True )<feature_engineering>
train_data["Fare_filled"] = train_data["Fare"].fillna(0) test_data["Fare_filled"] = test_data["Fare"].fillna(0 )
Titanic - Machine Learning from Disaster
13,055,333
ordem_porte = ['Pequeno porte 1', 'Pequeno porte 2', 'Médio porte', 'Grande porte'] df['porte'] = pd.Categorical(df['porte'], categories=ordem_porte ).codes df['div_gasto'] = df['gasto_pc_saude'] / df['pib'] df['exp_pib'] = df['pib'] / df['exp_vida'] df['exp_pib_estudos'] = df['pib'] / df['exp_anos_estudo']<data_type_conversions>
train_data["Fare_filled"] = train_data["Fare_filled"].astype(int) test_data["Fare_filled"] = test_data["Fare_filled"].astype(int )
Titanic - Machine Learning from Disaster
13,055,333
for col in df.columns: if df[col].dtype == 'object': df[col] = df[col].astype('category' ).cat.codes<split>
train_data["Cabin_filled"] = train_data["Cabin"].fillna("U0") deck = [] for data in train_data["Cabin_filled"]: if data[0][0] == "U": deck.append("U") elif data[0][0] == "A": deck.append("A") elif data[0][0] == "B": deck.append("B") elif data[0][0] == "C": deck.append("C") elif data[0][0] == "D": deck.append("D") elif data[0][0] == "E": deck.append("E") elif data[0][0] == "F": deck.append("F") elif data[0][0] == "G": deck.append("G") deck.append("U") train_data["Deck"] = deck
Titanic - Machine Learning from Disaster
13,055,333
test = df[df['nota_mat'].isnull() ] df = df[~df['nota_mat'].isnull() ]<filter>
test_data["Cabin_filled"] = test_data["Cabin"].fillna("U0") deck_test = [] for data in test_data["Cabin_filled"]: if data[0][0] == "U": deck_test.append("U") elif data[0][0] == "A": deck_test.append("A") elif data[0][0] == "B": deck_test.append("B") elif data[0][0] == "C": deck_test.append("C") elif data[0][0] == "D": deck_test.append("D") elif data[0][0] == "E": deck_test.append("E") elif data[0][0] == "F": deck_test.append("F") elif data[0][0] == "G": deck_test.append("G") test_data["Deck"] = deck_test
Titanic - Machine Learning from Disaster
13,055,333
df = df[df['populacao'] > 5000].copy()<split>
data = [train_data, test_data] for dataset in data: mean = train_data["Age"].mean() std = test_data["Age"].std() is_null = dataset["Age"].isnull().sum() rand_age = np.random.randint(mean - std, mean + std, size = is_null) age_slice = dataset["Age"].copy() age_slice[np.isnan(age_slice)] = rand_age dataset["Age_filled"] = age_slice dataset["Age_filled"] = dataset["Age_filled"].astype(int) test_data["Age_filled"].isnull().sum()
Titanic - Machine Learning from Disaster
13,055,333
train, valid = train_test_split(df, test_size=0.2, random_state=42 )<define_variables>
data = [train_data, test_data] for dataset in data: dataset['Age_filled'] = dataset['Age_filled'].astype(int) dataset.loc[ dataset['Age_filled'] <= 11, 'Age_filled'] = 0 dataset.loc[(dataset['Age_filled'] > 11)&(dataset['Age_filled'] <= 18), 'Age_filled'] = 1 dataset.loc[(dataset['Age_filled'] > 18)&(dataset['Age_filled'] <= 22), 'Age_filled'] = 2 dataset.loc[(dataset['Age_filled'] > 22)&(dataset['Age_filled'] <= 27), 'Age_filled'] = 3 dataset.loc[(dataset['Age_filled'] > 27)&(dataset['Age_filled'] <= 33), 'Age_filled'] = 4 dataset.loc[(dataset['Age_filled'] > 33)&(dataset['Age_filled'] <= 40), 'Age_filled'] = 5 dataset.loc[(dataset['Age_filled'] > 40)&(dataset['Age_filled'] <= 66), 'Age_filled'] = 6 dataset.loc[ dataset['Age_filled'] > 66, 'Age_filled'] = 6 test_data['Age_filled'].value_counts()
Titanic - Machine Learning from Disaster
13,055,333
feats = [c for c in df.columns if c not in ['nota_mat', 'codigo_mun', 'municipio']]<import_modules>
train_data["Deck_cat"] = train_data["Deck"].astype("category") test_data["Deck_cat"] = test_data["Deck"].astype("category" )
Titanic - Machine Learning from Disaster
13,055,333
from sklearn.ensemble import RandomForestClassifier<choose_model_class>
from sklearn.ensemble import RandomForestClassifier
Titanic - Machine Learning from Disaster
13,055,333
rf = RandomForestClassifier(n_estimators=200, min_samples_split=5, max_depth=4, random_state=42 )<train_model>
y = train_data["Survived"]
Titanic - Machine Learning from Disaster
13,055,333
rf.fit(train[feats], train['nota_mat'] )<predict_on_test>
X = train_data[["Pclass", "Sex_cat", "relatives", "Deck_cat", "Age_filled"]] X_test = test_data[["Pclass", "Sex_cat", "relatives", "Deck_cat", "Age_filled"]]
Titanic - Machine Learning from Disaster
13,055,333
preds = rf.predict(valid[feats] )<compute_test_metric>
X["Sex_cat"] = X["Sex_cat"].cat.codes X_test["Sex_cat"] = X_test["Sex_cat"].cat.codes
Titanic - Machine Learning from Disaster
13,055,333
accuracy_score(valid['nota_mat'], preds )<count_values>
X["Deck_cat"] = X["Deck_cat"].cat.codes X_test["Deck_cat"] = X_test["Deck_cat"].cat.codes
Titanic - Machine Learning from Disaster
13,055,333
valid['nota_mat'].value_counts()<predict_on_test>
model = RandomForestClassifier(n_estimators=100, max_depth=5, random_state=1)
Titanic - Machine Learning from Disaster
13,055,333
pred_test = rf.predict(test[feats] )<save_to_csv>
model.fit(X, y )
Titanic - Machine Learning from Disaster
13,055,333
r1 = pd.DataFrame({'codigo_mun': test['codigo_mun'], 'nota_mat': pred_test}) r1.to_csv('r1.csv', index=False )<compute_train_metric>
predictions = model.predict(X_test )
Titanic - Machine Learning from Disaster
13,055,333
def cv(df, test, feats, y_name, k=5): preds, score, fis = [], [], [] chunk = df.shape[0] // k for i in range(k): if i+1 < k: valid = df.iloc[i*chunk:(i+1)*chunk] train = df.iloc[:i*chunk].append(df.iloc[(i+1)*chunk:]) else: valid = df.iloc[i*chunk:] train = df.iloc[:i*chunk] rf = RandomForestClassifier(random_state=42, n_jobs=-1, n_estimators=200) rf.fit(train[feats], train[y_name]) score.append(accuracy_score(valid[y_name], rf.predict(valid[feats]))) preds.append(rf.predict(test[feats])) fis.append(rf.feature_importances_) print(i, 'OK') return score, preds, fis<compute_test_metric>
output = pd.DataFrame({'PassengerId': test_data.PassengerId, 'Survived': predictions} )
Titanic - Machine Learning from Disaster
13,055,333
score, preds, fis = cv(df, test, feats, 'nota_mat' )<save_to_csv>
output.to_csv('my_submission.csv', index=False) print("Your submission was successfully saved!" )
Titanic - Machine Learning from Disaster
13,290,484
r2 = pd.DataFrame({'codigo_mun': test['codigo_mun'], 'nota_mat': preds[1]}) r2.to_csv('r2.csv', index=False )<import_modules>
train_data = pd.read_csv(".. /input/titanic/train.csv") train_data.head()
Titanic - Machine Learning from Disaster
13,290,484
import lightgbm as lgb<compute_train_metric>
missing_val_count_by_column =(train_data.isnull().sum()) print(missing_val_count_by_column[missing_val_count_by_column > 0] )
Titanic - Machine Learning from Disaster
13,290,484
def cv(df, test, feats, y_name, k=5): preds, score, fis = [], [], [] chunk = df.shape[0] // k for i in range(k): if i+1 < k: valid = df.iloc[i*chunk:(i+1)*chunk] train = df.iloc[:i*chunk].append(df.iloc[(i+1)*chunk:]) else: valid = df.iloc[i*chunk:] train = df.iloc[:i*chunk] rf = lgb.LGBMClassifier(min_child_samples=500) rf.fit(train[feats], train[y_name], eval_set=[(train[feats], train['nota_mat']),(valid[feats], valid['nota_mat'])], early_stopping_rounds=50) score.append(accuracy_score(valid[y_name], rf.predict(valid[feats]))) preds.append(rf.predict(test[feats])) fis.append(rf.feature_importances_) print(i, 'OK') return score, preds, fis<compute_test_metric>
features = ['Pclass', 'Sex','SibSp','Age','Parch','Fare','Embarked'] X = pd.get_dummies(train_data[features]) y = train_data["Survived"]
Titanic - Machine Learning from Disaster
13,290,484
score, preds, fis = cv(df, test, feats, 'nota_mat' )<save_to_csv>
my_imputer = SimpleImputer() imputed_X = my_imputer.fit_transform(X )
Titanic - Machine Learning from Disaster
13,290,484
r3 = pd.DataFrame({'codigo_mun': test['codigo_mun'], 'nota_mat': preds[1]}) r3.to_csv('r3.csv', index=False )<set_options>
train_X, val_X, train_y, val_y = train_test_split(imputed_X,y,random_state=1 )
Titanic - Machine Learning from Disaster
13,290,484
gc.enable()<set_options>
tree = DecisionTreeClassifier() tree.fit(train_X, train_y) tree.score(val_X,val_y )
Titanic - Machine Learning from Disaster
13,290,484
tf.compat.v1.disable_eager_execution()<define_variables>
dumb_tree = DecisionTreeClassifier(max_depth=1) dumb_tree.fit(train_X, train_y) dumb_tree.score(val_X,val_y )
Titanic - Machine Learning from Disaster
13,290,484
CATEGORY_CNT = 54 BATCH_SIZE = 256 PREDICT_BATCH_SIZE = 256 EPOCHS = 10 STEPS_PER_EPOCH = 3000 VAL_STEPS_PER_EPOCH = 100 WORD2VEC_DIM = 200 FASTTEXT_DIM = 200 EXP_DECAY_COEF = 0.75 VERBOSE = 1 MAXLEN = 300 TRAINABLE_EMBED_SIZE = 75<load_pretrained>
smart_tree = DecisionTreeClassifier(max_depth=4,min_impurity_decrease=.004) smart_tree.fit(train_X, train_y) smart_tree.score(val_X,val_y )
Titanic - Machine Learning from Disaster
13,290,484
def save_object(obj, name): f = open(name, 'wb') pickle.dump(obj, f) f.close() def load_object(name): f = open(name, 'rb') obj = pickle.load(f) f.close() return obj<load_pretrained>
def get_score(max_leaf_nodes, train_X, val_X, train_y, val_y): model = DecisionTreeClassifier(max_leaf_nodes=max_leaf_nodes) model.fit(train_X, train_y) print("Max leaf nodes:", max_leaf_nodes," \t \t Score:", model.score(val_X,val_y)) return() for max_leaf_nodes in [2,3,4,5,6,7,8,9]: get_score(max_leaf_nodes, train_X, val_X, train_y, val_y )
Titanic - Machine Learning from Disaster
13,290,484
b_cat_map = load_object(".. /input/bonus-task-pickled-data-1/b_cat_map.pkl") f_cat_map = load_object(".. /input/bonus-task-pickled-data-1/f_cat_map.pkl") embed_fasttext = load_object(".. /input/bonus-task-pickled-data-1/embed_fasttext.pkl") embed_word2vec = load_object(".. /input/bonus-task-pickled-data-1/embed_word2vec.pkl") itemid = load_object(".. /input/bonus-task-pickled-data-1/itemid.pkl") y_train = load_object(".. /input/bonus-task-pickled-data-1/y_train.pkl") word_index = load_object(".. /input/bonus-x/word_index.pkl" )<categorify>
test_data = pd.read_csv(".. /input/titanic/test.csv") X_test=pd.get_dummies(test_data[features]) imputed_test=my_imputer.fit_transform(X_test )
Titanic - Machine Learning from Disaster