kernel_id int64 24.2k 23.3M | prompt stringlengths 8 1.85M | completetion stringlengths 1 182k | comp_name stringlengths 5 57 |
|---|---|---|---|
1,533,277 | df['category_id'] = df['Category'].factorize() [0]
df['category_id'][0:10]<remove_duplicates> | tmp = previous[previous['NAME_CONTRACT_STATUS'] != 'Approved'].groupby(['SK_ID_CURR'])['DAYS_DECISION']\
.agg({"returns": [np.min, np.max,np.mean]})\
.reset_index()
tmp.columns = ['SK_ID_CURR','des1','des2','des3']
for df in [train,test]:
tmp_merge = df[['SK_ID_CURR']]
tmp_merge = tmp_merge.merge(tmp, on=['SK_ID_CURR... | Home Credit Default Risk |
1,533,277 | category_id_df = df[['Category', 'category_id']].drop_duplicates().sort_values('category_id' )<define_variables> | tmp = previous[(~previous['NAME_CASH_LOAN_PURPOSE'].isin(['XAP','XNA'])) ]
for df in [train,test]:
tmp1 = tmp.groupby(['SK_ID_CURR'])['SK_ID_PREV'].count().reset_index()
tmp1.columns = ['SK_ID_CURR','des']
tmp_merge = df[['SK_ID_CURR']]
tmp_merge = tmp_merge.merge(tmp1, on=['SK_ID_CURR'], how='left')
df['count_clear_r... | Home Credit Default Risk |
1,533,277 | category_to_id = dict(category_id_df.values)
id_to_category = dict(category_id_df[['category_id', 'Category']].values )<count_values> | tmp = previous.groupby(['SK_ID_CURR'])['DAYS_TERMINATION']\
.agg({"returns": [np.min, np.max,np.mean]})\
.reset_index()
tmp.columns = ['SK_ID_CURR','des1','des2','des3']
for df in [train,test]:
tmp_merge = df[['SK_ID_CURR']]
tmp_merge = tmp_merge.merge(tmp, on=['SK_ID_CURR'], how='left')
df['min_day_termination'] = ... | Home Credit Default Risk |
1,533,277 | df.groupby('Category' ).category_id.count()
<categorify> | tmp = previous.groupby(['SK_ID_CURR'])['DAYS_LAST_DUE_1ST_VERSION']\
.agg({"returns": [np.min, np.max,np.mean]})\
.reset_index()
tmp.columns = ['SK_ID_CURR','des1','des2','des3']
for df in [train,test]:
tmp_merge = df[['SK_ID_CURR']]
tmp_merge = tmp_merge.merge(tmp, on=['SK_ID_CURR'], how='left')
df['min_day_lastdue... | Home Credit Default Risk |
1,533,277 | tfidf = TfidfVectorizer(sublinear_tf=True, min_df=5, norm='l2', encoding='latin-1', ngram_range=(1, 2), stop_words='english')
features = tfidf.fit_transform(df.Text ).toarray()
labels = df.category_id
<sort_values> | tmp = previous[~previous['DAYS_LAST_DUE_1ST_VERSION'].isnull() ].sort_values(by=['SK_ID_CURR','DAYS_LAST_DUE_1ST_VERSION'])
tmp = tmp.groupby(['SK_ID_CURR'] ).nth(-2 ).reset_index()
tmp = tmp[['SK_ID_CURR','DAYS_LAST_DUE_1ST_VERSION']]
tmp.columns = ['SK_ID_CURR','des']
for df in [train,test]:
tmp_merge = df[['SK_ID_C... | Home Credit Default Risk |
1,533,277 | sorted(category_to_id.items() )<statistical_test> | tmp = previous.groupby(['SK_ID_CURR'])['sooner']\
.agg({"returns": [np.min, np.max,np.mean]})\
.reset_index()
tmp.columns = ['SK_ID_CURR','des1','des2','des3']
for df in [train,test]:
tmp_merge = df[['SK_ID_CURR']]
tmp_merge = tmp_merge.merge(tmp, on=['SK_ID_CURR'], how='left')
df['min_sooner'] = tmp_merge['des1']
d... | Home Credit Default Risk |
1,533,277 | N = 3
for Category, category_id in sorted(category_to_id.items()):
features_chi2 = chi2(features, labels == category_id)
indices = np.argsort(features_chi2[0])
feature_names = np.array(tfidf.get_feature_names())[indices]
unigrams = [v for v in feature_names if len(v.split(' ')) == 1]
bigrams = [v for v in feature_nam... | tmp = previous.groupby(['SK_ID_CURR'])['SELLERPLACE_AREA']\
.agg({"returns": [np.min, np.max,np.mean]})\
.reset_index()
tmp.columns = ['SK_ID_CURR','des1','des2','des3']
for df in [train,test]:
tmp_merge = df[['SK_ID_CURR']]
tmp_merge = tmp_merge.merge(tmp, on=['SK_ID_CURR'], how='left')
df['min_seller'] = tmp_merge... | Home Credit Default Risk |
1,533,277 | SAMPLE_SIZE = int(len(features)* 0.3)
np.random.seed(0)
indices = np.random.choice(range(len(features)) , size=SAMPLE_SIZE, replace=False)
projected_features = TSNE(n_components=2, random_state=0 ).fit_transform(features[indices])
<filter> | for i in ['middle','low_normal','high','low_action']:
tmp = previous[(previous['NAME_CONTRACT_TYPE'] == 'Cash loans')&(previous['NAME_YIELD_GROUP'] == i)]
for df in [train,test]:
tmp1 = tmp.groupby(['SK_ID_CURR'])['SK_ID_PREV'].count().reset_index()
tmp1.columns = ['SK_ID_CURR','des']
tmp_merge = df[['SK_ID_CURR']]
tmp... | Home Credit Default Risk |
1,533,277 | my_id = 0
projected_features[(labels[indices] == my_id ).values]<choose_model_class> | for df in [train,test]:
df['tmp'] = df[['count_middle','count_low_normal','count_high','count_low_action']].sum(axis=1)
for i in ['middle','low_normal','high','low_action']:
df['ratio_' + i] = df['count_' + i]/df['tmp']
| Home Credit Default Risk |
1,533,277 | models = [
RandomForestClassifier(n_estimators=200, max_depth=3, random_state=0),
MultinomialNB() ,
LogisticRegression(random_state=0),
]
<create_dataframe> | for i in ['middle','low_normal','high','low_action']:
tmp = previous[(previous['NAME_CONTRACT_TYPE'] == 'Consumer loans')&(previous['NAME_YIELD_GROUP'] == i)]
for df in [train,test]:
tmp1 = tmp.groupby(['SK_ID_CURR'])['SK_ID_PREV'].count().reset_index()
tmp1.columns = ['SK_ID_CURR','des']
tmp_merge = df[['SK_ID_CURR']]... | Home Credit Default Risk |
1,533,277 | CV = 5
cv_df = pd.DataFrame(index=range(CV * len(models)))
entries = []<find_best_model_class> | for df in [train,test]:
df['tmp'] = df[['count_middle_v1','count_low_normal_v1','count_high_v1','count_low_action_v1']].sum(axis=1)
for i in ['middle','low_normal','high','low_action']:
df['ratio_' + i +"_v1"] = df['count_' + i + "_v1"]/df['tmp']
| Home Credit Default Risk |
1,533,277 | for model in models:
model_name = model.__class__.__name__
accuracies = cross_val_score(model, features, labels, scoring='accuracy', cv=CV)
for fold_idx, accuracy in enumerate(accuracies):
entries.append(( model_name, fold_idx, accuracy))<create_dataframe> | previous['tmp'] =(previous['AMT_ANNUITY'] * previous['CNT_PAYMENT'])/previous['AMT_CREDIT']
tmp = previous[(previous['NAME_CONTRACT_TYPE'] == 'Cash loans')&(previous['NAME_CONTRACT_STATUS'] != 'Approved')].groupby(['SK_ID_CURR'])['tmp']\
.agg({"returns": [np.min, np.max,np.mean]})\
.reset_index()
tmp.columns = ['SK_I... | Home Credit Default Risk |
1,533,277 | cv_df = pd.DataFrame(entries, columns=['model_name', 'fold_idx', 'accuracy'] )<groupby> | tmp = previous.groupby(['SK_ID_CURR'])['DAYS_FIRST_DRAWING']\
.agg({"returns": [np.min, np.max,np.mean]})\
.reset_index()
tmp.columns = ['SK_ID_CURR','des1','des2','des3']
for df in [train,test]:
tmp_merge = df[['SK_ID_CURR']]
tmp_merge = tmp_merge.merge(tmp, on=['SK_ID_CURR'], how='left')
df['min_firstdraw'] = tmp_... | Home Credit Default Risk |
1,533,277 | cv_df.groupby('model_name' ).accuracy.mean()<train_model> | previous['tmp'] = previous['DAYS_FIRST_DRAWING'] - previous['DAYS_DECISION']
tmp = previous.groupby(['SK_ID_CURR'])['tmp']\
.agg({"returns": [np.min, np.max,np.mean]})\
.reset_index()
tmp.columns = ['SK_ID_CURR','des1','des2','des3']
for df in [train,test]:
tmp_merge = df[['SK_ID_CURR']]
tmp_merge = tmp_merge.merge(t... | Home Credit Default Risk |
1,533,277 | model = LogisticRegression(random_state=0)
X_train, X_test, y_train, y_test, indices_train, indices_test = train_test_split(features, labels, df.index, test_size=0.33, random_state=0)
model.fit(X_train, y_train)
y_pred_proba = model.predict_proba(X_test)
y_pred = model.predict(X_test )<compute_test_metric> | previous['tmp'] = previous['DAYS_FIRST_DUE'] - previous['DAYS_FIRST_DRAWING']
tmp = previous.groupby(['SK_ID_CURR'])['tmp']\
.agg({"returns": [np.min, np.max,np.mean]})\
.reset_index()
tmp.columns = ['SK_ID_CURR','des1','des2','des3']
for df in [train,test]:
tmp_merge = df[['SK_ID_CURR']]
tmp_merge = tmp_merge.merge(... | Home Credit Default Risk |
1,533,277 | for predicted in category_id_df.category_id:
for actual in category_id_df.category_id:
if predicted != actual and conf_mat[actual, predicted] >= 2:
print("'{}' predicted as '{}' : {} examples.".format(id_to_category[actual], id_to_category[predicted], conf_mat[actual, predicted]))
display(df.loc[indices_test[(y_test ==... | previous['tmp'] = previous['DAYS_LAST_DUE'] - previous['DAYS_FIRST_DRAWING']
tmp = previous.groupby(['SK_ID_CURR'])['tmp']\
.agg({"returns": [np.min, np.max,np.mean]})\
.reset_index()
tmp.columns = ['SK_ID_CURR','des1','des2','des3']
for df in [train,test]:
tmp_merge = df[['SK_ID_CURR']]
tmp_merge = tmp_merge.merge(t... | Home Credit Default Risk |
1,533,277 | model.fit(features, labels )<features_selection> | tmp = bureau[(bureau['CREDIT_ACTIVE'] == "Active")].groupby(['SK_ID_CURR'])['SK_ID_BUREAU'].count().reset_index()
tmp.columns = ['SK_ID_CURR','des']
for df in [train,test]:
tmp_merge = df[['SK_ID_CURR']]
tmp_merge = tmp_merge.merge(tmp, on=['SK_ID_CURR'], how='left')
df['count_active_bureau'] = tmp_merge['des'].fillna... | Home Credit Default Risk |
1,533,277 | N = 5
for Category, category_id in sorted(category_to_id.items()):
indices = np.argsort(model.coef_[category_id])
feature_names = np.array(tfidf.get_feature_names())[indices]
unigrams = [v for v in reversed(feature_names)if len(v.split(' ')) == 1][:N]
bigrams = [v for v in reversed(feature_names)if len(v.split(' ')) =... | tmp = bureau[(bureau['CREDIT_TYPE'] == "Consumer credit")].groupby(['SK_ID_CURR'])['SK_ID_BUREAU'].count().reset_index()
tmp.columns = ['SK_ID_CURR','des']
for df in [train,test]:
tmp_merge = df[['SK_ID_CURR']]
tmp_merge = tmp_merge.merge(tmp, on=['SK_ID_CURR'], how='left')
df['count_active_bureau_v2'] = tmp_merge['de... | Home Credit Default Risk |
1,533,277 | texts = ["Hooli stock price soared after a dip in PiedPiper revenue growth.",
"Captain Tsubasa scores a magnificent goal for the Japanese team.",
"Merryweather mercenaries are sent on another mission, as government oversight groups call for new sanctions.",
"Beyoncé releases a new album, tops the charts in all of south... | tmp = bureau[(~bureau['CREDIT_TYPE'].isin(["Credit card","Consumer credit"])) ].groupby(['SK_ID_CURR'])['SK_ID_BUREAU'].count().reset_index()
tmp.columns = ['SK_ID_CURR','des']
for df in [train,test]:
tmp_merge = df[['SK_ID_CURR']]
tmp_merge = tmp_merge.merge(tmp, on=['SK_ID_CURR'], how='left')
df['count_active_bureau... | Home Credit Default Risk |
1,533,277 | TEST_PATH = os.path.join(".. /input/bbc-test", "BBC News Test.csv")
test_df = pd.read_csv(TEST_PATH)
<data_type_conversions> | bureau['AMT_CREDIT_SUM'].replace(0, np.nan, inplace = True)
tmp = bureau[(bureau['CREDIT_ACTIVE'] == "Active")&(bureau['CREDIT_TYPE'] == "Consumer credit")].groupby(['SK_ID_CURR'])['AMT_CREDIT_SUM'].agg({"returns": [np.min, np.max,np.mean, np.sum]})\
.reset_index()
tmp.columns = ['SK_ID_CURR','des1','des2','des3', 'd... | Home Credit Default Risk |
1,533,277 | test_df.Text.tolist()<predict_on_test> | bureau['AMT_CREDIT_SUM'].replace(0, np.nan, inplace = True)
tmp = bureau[(bureau['CREDIT_TYPE'] == "Credit card")].groupby(['SK_ID_CURR'])['AMT_CREDIT_SUM'].agg({"returns": [np.min, np.max,np.mean, np.sum]})\
.reset_index()
tmp.columns = ['SK_ID_CURR','des1','des2','des3', 'des4']
for df in [train,test]:
tmp_merge = ... | Home Credit Default Risk |
1,533,277 | test_features = tfidf.transform(test_df.Text.tolist())
Y_pred = model.predict(test_features )<define_variables> | bureau['AMT_CREDIT_SUM'].replace(0, np.nan, inplace = True)
tmp = bureau[(bureau['CREDIT_TYPE'] == "Car loan")].groupby(['SK_ID_CURR'])['AMT_CREDIT_SUM'].agg({"returns": [np.min, np.max,np.mean, np.sum]})\
.reset_index()
tmp.columns = ['SK_ID_CURR','des1','des2','des3', 'des4']
for df in [train,test]:
tmp_merge = df[... | Home Credit Default Risk |
1,533,277 | Y_pred_name =[]
for cat_id in Y_pred :
Y_pred_name.append(id_to_category[cat_id] )<create_dataframe> | bureau['AMT_CREDIT_SUM'].replace(0, np.nan, inplace = True)
tmp = bureau[(~bureau['CREDIT_TYPE'].isin(["Credit card","Consumer credit","Car loan"])) ].groupby(['SK_ID_CURR'])['AMT_CREDIT_SUM'].agg({"returns": [np.min, np.max,np.mean, np.sum]})\
.reset_index()
tmp.columns = ['SK_ID_CURR','des1','des2','des3', 'des4']
... | Home Credit Default Risk |
1,533,277 | submission = pd.DataFrame({
"ArticleId": test_df["ArticleId"],
"Category": Y_pred_name
} )<save_to_csv> | tmp = bureau.groupby(['SK_ID_CURR'])['AMT_CREDIT_SUM'].agg({"returns": [np.min, np.max,np.mean, np.sum]})\
.reset_index()
tmp.columns = ['SK_ID_CURR','des1','des2','des3', 'des4']
for df in [train,test]:
tmp_merge = df[['SK_ID_CURR']]
tmp_merge = tmp_merge.merge(tmp, on=['SK_ID_CURR'], how='left')
df['min_credit_bure... | Home Credit Default Risk |
1,533,277 | submission.to_csv('submission.csv', index=False )<load_from_csv> | tmp = bureau[(bureau['CREDIT_TYPE'] == "Consumer credit")].groupby(['SK_ID_CURR'])['DAYS_CREDIT_ENDDATE'].agg({"returns": [np.min, np.max,np.mean, np.sum]})\
.reset_index()
tmp.columns = ['SK_ID_CURR','des1','des2','des3', 'des4']
for df in [train,test]:
tmp_merge = df[['SK_ID_CURR']]
tmp_merge = tmp_merge.merge(tmp, ... | Home Credit Default Risk |
1,533,277 | TRAIN_PATH = os.path.join(".. /input/ai-academy-intermediate-class-competition-1", "BBC News Train.csv")
df = pd.read_csv(TRAIN_PATH )<feature_engineering> | tmp = bureau[(bureau['CREDIT_TYPE'] == "Car loan")].groupby(['SK_ID_CURR'])['DAYS_CREDIT_ENDDATE'].agg({"returns": [np.min, np.max,np.mean, np.sum]})\
.reset_index()
tmp.columns = ['SK_ID_CURR','des1','des2','des3', 'des4']
for df in [train,test]:
tmp_merge = df[['SK_ID_CURR']]
tmp_merge = tmp_merge.merge(tmp, on=['SK... | Home Credit Default Risk |
1,533,277 | df['category_id'] = df['Category'].factorize() [0]
df['category_id'][0:10]<remove_duplicates> | tmp = bureau[(bureau['CREDIT_TYPE'] == "Consumer credit")].groupby(['SK_ID_CURR'])['DAYS_CREDIT'].agg({"returns": [np.min, np.max,np.mean, np.sum]})\
.reset_index()
tmp.columns = ['SK_ID_CURR','des1','des2','des3', 'des4']
for df in [train,test]:
tmp_merge = df[['SK_ID_CURR']]
tmp_merge = tmp_merge.merge(tmp, on=['SK_... | Home Credit Default Risk |
1,533,277 | category_id_df = df[['Category', 'category_id']].drop_duplicates().sort_values('category_id' )<define_variables> | tmp = bureau[(~bureau['DAYS_CREDIT_ENDDATE'].isnull())&(( bureau['CREDIT_TYPE'] == "Consumer credit")) ].sort_values(by=['SK_ID_CURR','DAYS_CREDIT_ENDDATE'])
tmp = tmp.groupby(['SK_ID_CURR'] ).nth(-2 ).reset_index()
tmp = tmp[['SK_ID_CURR','DAYS_ENDDATE_FACT']]
tmp.columns = ['SK_ID_CURR','des']
for df in [train,test]... | Home Credit Default Risk |
1,533,277 | category_to_id = dict(category_id_df.values)
id_to_category = dict(category_id_df[['category_id', 'Category']].values )<count_values> | tmp = bureau[(bureau['CREDIT_TYPE'] == "Consumer credit")].groupby(['SK_ID_CURR'])['DAYS_ENDDATE_FACT'].agg({"returns": [np.min, np.max,np.mean, np.sum]})\
.reset_index()
tmp.columns = ['SK_ID_CURR','des1','des2','des3', 'des4']
for df in [train,test]:
tmp_merge = df[['SK_ID_CURR']]
tmp_merge = tmp_merge.merge(tmp, on... | Home Credit Default Risk |
1,533,277 | df.groupby('Category' ).category_id.count()
<categorify> | bureau['tmp'] = bureau['DAYS_CREDIT_ENDDATE'] - bureau['DAYS_ENDDATE_FACT']
tmp = bureau.groupby(['SK_ID_CURR'])['tmp'].agg({"returns": [np.min, np.max,np.mean, np.sum]})\
.reset_index()
tmp.columns = ['SK_ID_CURR','des1','des2','des3', 'des4']
for df in [train,test]:
tmp_merge = df[['SK_ID_CURR']]
tmp_merge = tmp_mer... | Home Credit Default Risk |
1,533,277 | tfidf = TfidfVectorizer(sublinear_tf=True, min_df=5, norm='l2', encoding='latin-1', ngram_range=(1, 2), stop_words='english')
features = tfidf.fit_transform(df.Text ).toarray()
labels = df.category_id
<sort_values> | bureau['tmp'] =(bureau['DAYS_CREDIT_ENDDATE'] - bureau['DAYS_CREDIT'])
tmp = bureau[(bureau['CREDIT_TYPE'] == "Consumer credit")].groupby(['SK_ID_CURR'])['tmp'].agg({"returns": [np.min, np.max,np.mean, np.sum]})\
.reset_index()
tmp.columns = ['SK_ID_CURR','des1','des2','des3', 'des4']
for df in [train,test]:
tmp_merg... | Home Credit Default Risk |
1,533,277 | sorted(category_to_id.items() )<statistical_test> | bureau['tmp'] =(bureau['DAYS_ENDDATE_FACT'] - bureau['DAYS_CREDIT'])
tmp = bureau[(bureau['CREDIT_TYPE'] == "Consumer credit")].groupby(['SK_ID_CURR'])['tmp'].agg({"returns": [np.min, np.max,np.mean, np.sum]})\
.reset_index()
tmp.columns = ['SK_ID_CURR','des1','des2','des3', 'des4']
for df in [train,test]:
tmp_merge ... | Home Credit Default Risk |
1,533,277 | N = 3
for Category, category_id in sorted(category_to_id.items()):
features_chi2 = chi2(features, labels == category_id)
indices = np.argsort(features_chi2[0])
feature_names = np.array(tfidf.get_feature_names())[indices]
unigrams = [v for v in feature_names if len(v.split(' ')) == 1]
bigrams = [v for v in feature_nam... | bureau['tmp'] =(bureau['DAYS_ENDDATE_FACT'] - bureau['DAYS_CREDIT_ENDDATE'])/(bureau['DAYS_CREDIT'] - bureau['DAYS_CREDIT_ENDDATE'])
tmp = bureau[(bureau['CREDIT_TYPE'] == "Consumer credit")].groupby(['SK_ID_CURR'])['tmp'].agg({"returns": [np.min, np.max,np.mean, np.sum]})\
.reset_index()
tmp.columns = ['SK_ID_CURR',... | Home Credit Default Risk |
1,533,277 | SAMPLE_SIZE = int(len(features)* 0.3)
np.random.seed(0)
indices = np.random.choice(range(len(features)) , size=SAMPLE_SIZE, replace=False)
projected_features = TSNE(n_components=2, random_state=0 ).fit_transform(features[indices])
<filter> | bureau['tmp'] =(bureau['DAYS_ENDDATE_FACT'] - bureau['DAYS_CREDIT_ENDDATE'])/(bureau['DAYS_CREDIT'] - bureau['DAYS_CREDIT_ENDDATE'])
tmp = bureau[(~bureau['CREDIT_TYPE'].isin(['Credit card','Consumer credit'])) ].groupby(['SK_ID_CURR'])['tmp'].agg({"returns": [np.min, np.max,np.mean, np.sum]})\
.reset_index()
tmp.col... | Home Credit Default Risk |
1,533,277 | my_id = 0
projected_features[(labels[indices] == my_id ).values]<choose_model_class> | bureau['tmp'] =(bureau['AMT_CREDIT_SUM'])/(bureau['DAYS_CREDIT'] - bureau['DAYS_CREDIT_ENDDATE'])
tmp = bureau[bureau['CREDIT_TYPE'] == "Credit card"].groupby(['SK_ID_CURR'])['tmp'].agg({"returns": [np.min, np.max,np.mean, np.sum]})\
.reset_index()
tmp.columns = ['SK_ID_CURR','des1','des2','des3', 'des4']
for df in [... | Home Credit Default Risk |
1,533,277 | models = [
RandomForestClassifier(n_estimators=200, max_depth=3, random_state=0),
MultinomialNB() ,
LogisticRegression(random_state=0),
]
<create_dataframe> | bureau['tmp'] =(bureau['AMT_CREDIT_SUM'])/(bureau['DAYS_CREDIT'] - bureau['DAYS_CREDIT_ENDDATE'])
tmp = bureau[(bureau['CREDIT_ACTIVE'] == "Active")&(bureau['CREDIT_TYPE'] == "Consumer credit")].groupby(['SK_ID_CURR'])['tmp'].agg({"returns": [np.min, np.max,np.mean, np.sum]})\
.reset_index()
tmp.columns = ['SK_ID_CUR... | Home Credit Default Risk |
1,533,277 | CV = 5
cv_df = pd.DataFrame(index=range(CV * len(models)))
entries = []<find_best_model_class> | bureau['tmp'] =(bureau['AMT_CREDIT_SUM'])/(bureau['DAYS_CREDIT'] - bureau['DAYS_CREDIT_ENDDATE'])
tmp = bureau[~bureau['CREDIT_TYPE'].isin(["Credit card","Consumer credit"])].groupby(['SK_ID_CURR'])['tmp'].agg({"returns": [np.min, np.max,np.mean, np.sum]})\
.reset_index()
tmp.columns = ['SK_ID_CURR','des1','des2','de... | Home Credit Default Risk |
1,533,277 | for model in models:
model_name = model.__class__.__name__
accuracies = cross_val_score(model, features, labels, scoring='accuracy', cv=CV)
for fold_idx, accuracy in enumerate(accuracies):
entries.append(( model_name, fold_idx, accuracy))<create_dataframe> | bureau['AMT_CREDIT_SUM_DEBT_v1'] = bureau['AMT_CREDIT_SUM_DEBT'].replace(0, np.nan)
tmp = bureau[(bureau['CREDIT_ACTIVE'] == "Active")&(bureau['CREDIT_TYPE'].isin(["Credit card"])) ].groupby(['SK_ID_CURR'])['AMT_CREDIT_SUM_DEBT_v1'].agg({"returns": [np.min, np.max,np.mean, np.sum]})\
.reset_index()
tmp.columns = ['SK... | Home Credit Default Risk |
1,533,277 | cv_df = pd.DataFrame(entries, columns=['model_name', 'fold_idx', 'accuracy'] )<groupby> | bureau['AMT_CREDIT_SUM_LIMIT_v1'] = bureau['AMT_CREDIT_SUM_LIMIT'].replace(0, np.nan)
tmp = bureau[(bureau['CREDIT_TYPE'].isin(["Credit card"])) ].groupby(['SK_ID_CURR'])['AMT_CREDIT_SUM_LIMIT_v1'].agg({"returns": [np.min, np.max,np.mean, np.sum]})\
.reset_index()
tmp.columns = ['SK_ID_CURR','des1','des2','des3', 'de... | Home Credit Default Risk |
1,533,277 | cv_df.groupby('model_name' ).accuracy.mean()<train_model> | bureau['AMT_CREDIT_MAX_OVERDUE_v1'] = bureau['AMT_CREDIT_MAX_OVERDUE'].replace(0,np.nan)
tmp = bureau[(bureau['CREDIT_TYPE'].isin(["Consumer credit"])) ].groupby(['SK_ID_CURR'])['AMT_CREDIT_MAX_OVERDUE_v1'].agg({"returns": [np.min, np.max,np.mean, np.sum]})\
.reset_index()
tmp.columns = ['SK_ID_CURR','des1','des2','d... | Home Credit Default Risk |
1,533,277 | model = LogisticRegression(random_state=0)
X_train, X_test, y_train, y_test, indices_train, indices_test = train_test_split(features, labels, df.index, test_size=0.33, random_state=0)
model.fit(X_train, y_train)
y_pred_proba = model.predict_proba(X_test)
y_pred = model.predict(X_test )<compute_test_metric> | bureau['tmp'] = bureau['AMT_CREDIT_SUM_DEBT']/bureau['AMT_CREDIT_SUM']
tmp = bureau.groupby(['SK_ID_CURR'])['tmp'].agg({"returns": [np.min, np.max,np.mean, np.sum]})\
.reset_index()
tmp.columns = ['SK_ID_CURR','des1','des2','des3', 'des4']
for df in [train,test]:
tmp_merge = df[['SK_ID_CURR']]
tmp_merge = tmp_merge.me... | Home Credit Default Risk |
1,533,277 | for predicted in category_id_df.category_id:
for actual in category_id_df.category_id:
if predicted != actual and conf_mat[actual, predicted] >= 2:
print("'{}' predicted as '{}' : {} examples.".format(id_to_category[actual], id_to_category[predicted], conf_mat[actual, predicted]))
display(df.loc[indices_test[(y_test ==... | bureau['tmp'] = bureau['AMT_ANNUITY'].fillna(0)
tmp = bureau.groupby(['SK_ID_CURR'])['tmp'].agg({"returns": [np.min, np.max,np.mean, np.sum]})\
.reset_index()
tmp.columns = ['SK_ID_CURR','des1','des2','des3', 'des4']
for df in [train,test]:
tmp_merge = df[['SK_ID_CURR']]
tmp_merge = tmp_merge.merge(tmp, on=['SK_ID_CU... | Home Credit Default Risk |
1,533,277 | model.fit(features, labels )<features_selection> | install = pd.read_csv(".. /input/installments_payments.csv" ) | Home Credit Default Risk |
1,533,277 | N = 5
for Category, category_id in sorted(category_to_id.items()):
indices = np.argsort(model.coef_[category_id])
feature_names = np.array(tfidf.get_feature_names())[indices]
unigrams = [v for v in reversed(feature_names)if len(v.split(' ')) == 1][:N]
bigrams = [v for v in reversed(feature_names)if len(v.split(' ')) =... | tmp = install.groupby(['SK_ID_PREV','SK_ID_CURR'] ).agg({'NUM_INSTALMENT_NUMBER':["count","max"]} ).reset_index()
tmp.columns = ['SK_ID_PREV','SK_ID_CURR','count','max']
tmp['delta'] = tmp['count'] - tmp['max']
tmp = tmp.merge(previous[['SK_ID_PREV','CNT_PAYMENT','DAYS_LAST_DUE','NAME_CONTRACT_TYPE']], on=['SK_ID_PREV'... | Home Credit Default Risk |
1,533,277 | texts = ["Hooli stock price soared after a dip in PiedPiper revenue growth.",
"Captain Tsubasa scores a magnificent goal for the Japanese team.",
"Merryweather mercenaries are sent on another mission, as government oversight groups call for new sanctions.",
"Beyoncé releases a new album, tops the charts in all of south... | tmp = install.groupby(['SK_ID_PREV','SK_ID_CURR'] ).agg({'NUM_INSTALMENT_VERSION':["count","max"]} ).reset_index()
tmp.columns = ['SK_ID_PREV','SK_ID_CURR','count','max']
tmp = tmp.merge(previous[['SK_ID_PREV','CNT_PAYMENT','DAYS_LAST_DUE','NAME_CONTRACT_TYPE']], on=['SK_ID_PREV'], how='left')
tmp_1 = tmp.groupby(['SK... | Home Credit Default Risk |
1,533,277 | TEST_PATH = os.path.join(".. /input/bbc-test", "BBC News Test.csv")
test_df = pd.read_csv(TEST_PATH)
<predict_on_test> | tmp = install.groupby(['SK_ID_PREV','SK_ID_CURR'] ).agg({'NUM_INSTALMENT_NUMBER':["count","max"]} ).reset_index()
tmp.columns = ['SK_ID_PREV','SK_ID_CURR','count','max']
tmp['delta'] = tmp['count']/tmp['max']
tmp = tmp.merge(previous[['SK_ID_PREV','CNT_PAYMENT','DAYS_LAST_DUE','NAME_CONTRACT_TYPE']], on=['SK_ID_PREV'],... | Home Credit Default Risk |
1,533,277 | test_features = tfidf.transform(test_df.Text.tolist())
Y_pred = model.predict(test_features)
Y_pred<define_variables> | tmp = install[['SK_ID_PREV','SK_ID_CURR','NUM_INSTALMENT_NUMBER','AMT_INSTALMENT']].drop_duplicates() | Home Credit Default Risk |
1,533,277 | Y_pred_name =[]
for cat_id in Y_pred :
Y_pred_name.append(id_to_category[cat_id] )<create_dataframe> | tmp = tmp.groupby(['SK_ID_PREV','SK_ID_CURR'])['AMT_INSTALMENT'].sum().reset_index()
tmp.columns = ['SK_ID_PREV','SK_ID_CURR','need_to_pay'] | Home Credit Default Risk |
1,533,277 | submission = pd.DataFrame({
"ArticleId": test_df["ArticleId"],
"Category": Y_pred_name
} )<save_to_csv> | tmp_1 = install.groupby(['SK_ID_PREV'])['AMT_PAYMENT'].sum().reset_index()
tmp_1.columns = ['SK_ID_PREV','paid'] | Home Credit Default Risk |
1,533,277 | submission.to_csv('submission.csv', index=False )<set_options> | tmp = tmp.merge(tmp_1, on=['SK_ID_PREV'], how='left' ) | Home Credit Default Risk |
1,533,277 | %matplotlib inline
try:
except ImportError as e:
print('scikit-image is too new, ',e)
<set_options> | payment_history = tmp
payment_history['ratio'] = payment_history['paid']/payment_history['need_to_pay']
payment_history['delta'] = payment_history['need_to_pay'] - payment_history['paid']
payment_history = payment_history.merge(previous[['SK_ID_PREV','AMT_ANNUITY','CNT_PAYMENT','NAME_CONTRACT_TYPE']], \
on = ['SK_ID_PR... | Home Credit Default Risk |
1,533,277 | use_cuda = True
device = torch.device("cuda" if use_cuda else "cpu")
torch.manual_seed(42 )<train_model> | tmp = payment_history[payment_history['mean_version'] > 0].groupby(['SK_ID_CURR'])['ratio'].agg({"returns": [np.min, np.max,np.mean, np.sum]})\
.reset_index()
tmp.columns = ['SK_ID_CURR','des1','des2','des3', 'des4']
for df in [train,test]:
tmp_merge = df[['SK_ID_CURR']]
tmp_merge = tmp_merge.merge(tmp, on=['SK_ID_CUR... | Home Credit Default Risk |
1,533,277 | class AffMNISTDataset(Dataset):
def __init__(self, data_name):
with np.load(os.path.join(DATA_ROOT_PATH, '{}.npz'.format(data_name)))as npz_data:
self.img_vec = npz_data['img']
self.idx_vec = npz_data['idx']
print('image shape', self.img_vec.shape)
print('idx shape', self.idx_vec.shape)
label_path = os.path.join(DA... | tmp = payment_history[payment_history['mean_version'] > 0].groupby(['SK_ID_CURR'])['delta'].agg({"returns": [np.min, np.max,np.mean, np.sum]})\
.reset_index()
tmp.columns = ['SK_ID_CURR','des1','des2','des3', 'des4']
for df in [train,test]:
tmp_merge = df[['SK_ID_CURR']]
tmp_merge = tmp_merge.merge(tmp, on=['SK_ID_CUR... | Home Credit Default Risk |
1,533,277 | class SimpleMLP(nn.Module):
def __init__(self):
super(SimpleMLP, self ).__init__()
self.fc1 = nn.Linear(40*40, 512)
self.fc2 = nn.Linear(512, 10)
def forward(self, x):
x = x.view(-1, 40*40)
x = F.relu(self.fc1(x))
x = self.fc2(x)
return F.log_softmax(x, dim=1 )<create_dataframe> | tmp = install.groupby(['SK_ID_PREV','SK_ID_CURR'] ).agg({'NUM_INSTALMENT_NUMBER':["count","max"]} ).reset_index()
tmp.columns = ['SK_ID_PREV','SK_ID_CURR','count','max']
tmp['delta'] = tmp['count']/tmp['max']
tmp = tmp.merge(previous[['SK_ID_PREV','CNT_PAYMENT','DAYS_LAST_DUE','NAME_CONTRACT_TYPE']], on=['SK_ID_PREV'],... | Home Credit Default Risk |
1,533,277 | train_ds = AffMNISTDataset('train')
train_loader = DataLoader(train_ds, batch_size=1024,
shuffle=True, num_workers=4 )<train_model> | tmp = install.groupby(['SK_ID_PREV'])['AMT_PAYMENT'].max().reset_index()
tmp.columns = ['SK_ID_PREV','max_install']
payment_history = payment_history.merge(tmp, on=['SK_ID_PREV'], how='left' ) | Home Credit Default Risk |
1,533,277 | model = SimpleMLP().to(device)
model.train()<choose_model_class> | payment_history['tmp'] = payment_history['max_install']/payment_history['AMT_ANNUITY']
tmp = payment_history[payment_history['mean_version'] > 0].groupby(['SK_ID_CURR'])['tmp'] | Home Credit Default Risk |
1,533,277 | optimizer = optim.SGD(model.parameters() ,
lr=1e-3 )<train_model> | tmp = install.groupby(['SK_ID_PREV'])['NUM_INSTALMENT_NUMBER'].max().reset_index()
tmp.columns = ['SK_ID_PREV','max_num_install']
payment_history = payment_history.merge(tmp, on=['SK_ID_PREV'], how='left' ) | Home Credit Default Risk |
1,533,277 | log_interval = 100
for epoch in range(1):
for batch_idx,(data, target)in enumerate(train_loader):
data, target = data.to(device), target.to(device)
optimizer.zero_grad()
output = model(data)
loss = F.nll_loss(output, target)
loss.backward()
optimizer.step()
if batch_idx % log_interval == 0:
print('Train Epoch: {} [{... | tmp = install[install['AMT_INSTALMENT'] > install['AMT_PAYMENT']]
tmp = tmp.groupby(['SK_ID_CURR'])['SK_ID_PREV'].count().reset_index()
tmp.columns = ['SK_ID_CURR','des']
for df in [train,test]:
tmp_merge = df[['SK_ID_CURR']]
tmp_merge = tmp_merge.merge(tmp, on=['SK_ID_CURR'], how='left')
df['count_small_payment'] = t... | Home Credit Default Risk |
1,533,277 | test_ds = AffMNISTDataset('test')
test_loader = DataLoader(test_ds, batch_size=1024,
shuffle=True, num_workers=4 )<categorify> | install['tmp'] = install['AMT_PAYMENT']/install['AMT_INSTALMENT']
for i in range(10):
print(i)
tmp = install[(install['tmp'] > i/10)&(install['tmp'] <(( i+1)/10)) ]
tmp = tmp.groupby(['SK_ID_CURR'])['SK_ID_PREV'].count().reset_index()
tmp.columns = ['SK_ID_CURR','des']
for df in [train,test]:
tmp_merge = df[['SK_ID_CU... | Home Credit Default Risk |
1,533,277 | target_out = []
pred_out = []
for batch_idx,(data, target_idx)in enumerate(test_loader):
data = data.to(device)
target_idx = target_idx.to('cpu' ).numpy()
output = model(data)
pred = output.to('cpu' ).data.max(1)[1].numpy()
target_out += [target_idx]
pred_out += [pred]<save_to_csv> | tmp = install.groupby(['SK_ID_PREV','NUM_INSTALMENT_NUMBER'])['DAYS_INSTALMENT'].count().reset_index()
tmp = tmp[tmp['DAYS_INSTALMENT'] > 1]
tmp.columns = ['SK_ID_PREV','NUM_INSTALMENT_NUMBER','count_dup']
install = install.merge(tmp, on = ['SK_ID_PREV','NUM_INSTALMENT_NUMBER'], how='left')
dup_install = install[insta... | Home Credit Default Risk |
1,533,277 | pred_df = pd.DataFrame({'idx': np.concatenate(target_out, 0),
'label': np.concatenate(pred_out, 0)})
pred_df.to_csv('mlp_predictions.csv', index=False)
pred_df.sample(5 )<set_options> | tmp = install[(install['AMT_PAYMENT'] < install['AMT_INSTALMENT'])&(install['DAYS_ENTRY_PAYMENT'] < install['DAYS_INSTALMENT'])]
tmp['ratio'] = tmp['AMT_PAYMENT']/tmp['AMT_INSTALMENT']
tmp = dup_install.groupby(['SK_ID_CURR'])['AMT_PAYMENT'].agg({"returns": [np.min, np.max,np.mean, np.sum]})\
.reset_index()
tmp_1.colu... | Home Credit Default Risk |
1,533,277 | sns.set_context("paper")
print(tf.__version__ )<load_from_csv> | dup_install.sort_values(by=['SK_ID_PREV','NUM_INSTALMENT_NUMBER','DAYS_ENTRY_PAYMENT'] ) | Home Credit Default Risk |
1,533,277 | typetable = pd.read_csv(".. /input/pokemon-type-table/typetable.csv")
vals = []
for c1 in typetable.columns[1:]:
vals.append(pd.DataFrame({
"idx": typetable["atck"].map(lambda x: "%s-vs-%s-None" %(x, c1)) ,
"mul": typetable[c1],
}))
for c2 in typetable.columns[1:]:
vals.append(pd.DataFrame({
"idx": typetable["atck"].m... | credit = pd.read_csv(".. /input/credit_card_balance.csv" ) | Home Credit Default Risk |
1,533,277 | class ModelCallback(keras.callbacks.Callback):
def set_params(self, params):
self.epochs = params["epochs"]
def on_epoch_end(self, epoch, epoch_logs):
print("\r", "%5s/%-5s " %(epoch + 1, self.epochs), end="")
for k, v in epoch_logs.items() :
print("%s: %04.4f "%(k, v), end="" )<load_from_csv> | credit['tmp'] = credit['AMT_BALANCE']/credit['AMT_CREDIT_LIMIT_ACTUAL']
tmp = credit.groupby(["SK_ID_CURR","SK_ID_PREV"])['tmp'].max().reset_index()
tmp = tmp.groupby(['SK_ID_CURR'])['tmp'].agg({"returns": [np.min, np.max,np.mean, np.sum]})\
.reset_index()
tmp.columns = ['SK_ID_CURR','des1','des2','des3', 'des4']
for ... | Home Credit Default Risk |
1,533,277 | to_underscore = lambda x: re.sub("[^0-9a-zA-Z
load_csv = lambda file: pd.read_csv(".. /input/pokemon-challenge-mlh/%s.csv" % file ).rename(to_underscore, axis='columns')
to_underscore("First Colum.Value" )<feature_engineering> | credit['tmp'] = credit['AMT_BALANCE']/credit['AMT_CREDIT_LIMIT_ACTUAL']
tmp = credit.groupby(["SK_ID_CURR","SK_ID_PREV"])['tmp'].min().reset_index()
tmp = credit.groupby(['SK_ID_CURR'])['tmp'].agg({"returns": [np.min, np.max,np.mean, np.sum]})\
.reset_index()
tmp.columns = ['SK_ID_CURR','des1','des2','des3', 'des4']
f... | Home Credit Default Risk |
1,533,277 | def df_pokemon() :
pokemon = load_csv("pokemon" ).fillna("None")
pokemon = pokemon.drop(["name", "generation", "legendary"], axis = 1)
pokemon["speed2"] = pokemon["speed"] ** 2
pokemon["speed2"] = pokemon["speed2"] / pokemon["speed2"].max()
pokemon["speed"] = pokemon["speed"] / pokemon["speed"].max()
pokemon["hp"] = ... | doc = [x for x in train.columns if 'FLAG_DOC' in x]
connection = ['FLAG_MOBIL', 'FLAG_EMP_PHONE', 'FLAG_WORK_PHONE',
'FLAG_CONT_MOBILE', 'FLAG_PHONE', 'FLAG_EMAIL',]
le = LabelEncoder()
categorical = ['CODE_GENDER','FLAG_OWN_CAR','FLAG_OWN_REALTY','NAME_EDUCATION_TYPE',
'NAME_FAMILY_STATUS',
'FLAG_MOBIL','FLAG_EMP_PHON... | Home Credit Default Risk |
1,533,277 | def merge_data(battles, pokemon):
battles = battles \
.merge(pokemon.rename(lambda x: "f_%s" % x, axis="columns"), left_on="first_pokemon", right_on="f_
.merge(pokemon.rename(lambda x: "s_%s" % x, axis="columns"), left_on="second_pokemon", right_on="s_
battles["f_t1"] =(battles.f_type_1 + "-vs-" + battles.s_type_1 + ... | NFOLDS = 5
kf = StratifiedKFold(n_splits=NFOLDS, shuffle=True, random_state=2018)
pred_test_full = 0
params = {
'boosting': 'gbdt',
'objective': 'binary',
'metric': 'auc',
'learning_rate': 0.01,
'num_leaves': 40,
'max_depth': 7,
'colsample_bytree': 0.15,
'seed': 101
}
res = []
idx = 0
for dev_index, val_index in kf.sp... | Home Credit Default Risk |
1,533,277 | def params(train, units=[]):
X, Xs = to_model(train)
Y = train["winner"].values
y = keras.utils.to_categorical(Y)
assert X.shape[0] == y.shape[0]
input_dim = X.shape[1]
outuput_dim = 1 if len(y.shape)== 1 else y.shape[1]
samples = train.shape[0]
print("𝑁𝑖: %s, 𝑁𝑜: %s, 𝑁𝑠: %s" %(input_dim, outuput_dim, samples))... | tmp = install.groupby(['SK_ID_PREV','NUM_INSTALMENT_NUMBER'])['DAYS_INSTALMENT'].count().reset_index()
tmp = tmp[tmp['DAYS_INSTALMENT'] > 1]
tmp.columns = ['SK_ID_PREV','NUM_INSTALMENT_NUMBER','count_dup']
install = install.merge(tmp, on = ['SK_ID_PREV','NUM_INSTALMENT_NUMBER'], how='left')
tmp = install[install['coun... | Home Credit Default Risk |
1,533,277 | def to_model(X):
if "winner" in X:
X = X.drop(["winner"], axis=1)
X = X.values
return X, X<count_duplicates> | install.drop(['count_dup_x','count_dup_y'], axis=1, inplace = True ) | Home Credit Default Risk |
1,533,277 | def remove_duplicate_battles(battles):
return battles\
.groupby(["first_pokemon", "second_pokemon", "winner"])\
.count() \
.reset_index() \
.drop(["battle_number"],axis=1)\
.sample(frac=1)\
.reset_index(drop=True)\
.reset_index() \
.rename(columns={"index": "battle_number"} )<load_from_csv> | tmp.sort_values(by=['SK_ID_PREV','NUM_INSTALMENT_NUMBER','DAYS_ENTRY_PAYMENT'] ) | Home Credit Default Risk |
1,533,277 | battles_values = load_csv("battles" ).values
acc100 = dict(zip(map(tuple, battles_values[:, 1:-1]), map(tuple, battles_values[:, -1:])) )<train_model> | tmp['SK_ID_CURR'].value_counts() | Home Credit Default Risk |
1,316,642 | epochs, validation_split =(600, 0.0)
history = model.fit(Xs, y, workers=4, epochs=epochs, verbose=0, validation_split=validation_split,
batch_size=512, shuffle=True, callbacks=[ModelCallback() ], initial_epoch=0)
plot_history(( history), title="", since=0 )<save_to_csv> | import matplotlib.pyplot as plt
import lightgbm as lgb
import gc
from sklearn.model_selection import KFold, cross_val_score
from sklearn.metrics import confusion_matrix,precision_recall_curve,auc,roc_auc_score,roc_curve,recall_score,classification_report
from sklearn.preprocessing import LabelEncoder | Home Credit Default Risk |
1,316,642 | test = load_csv("test")
X, Xs = to_model(merge_data(super_battle(load_csv("test")) , pokemon))
prediction = model.predict_classes(Xs)
df_submission = pd.DataFrame({"Winner": prediction}, index = test.index.rename("battle_number"))
for i, p1, p2 in load_csv("test" ).values:
if(p1, p2)in acc100: df_submission.iloc[i] =... | app_train = pd.read_csv('.. /input/application_train.csv')
print('Training data shape: ', app_train.shape)
app_train.head() | Home Credit Default Risk |
1,316,642 | model.save_weights("cp-{epoch:04d}.ckpt".format(epoch=epochs))
! mkdir "saved_models"
saved_model_path = tf.contrib.saved_model.save_keras_model(model, "./saved_models" )<load_from_csv> | app_test = pd.read_csv('.. /input/application_test.csv')
print('Testing data shape: ', app_test.shape)
app_test.head() | Home Credit Default Risk |
1,316,642 | battles = pd.read_csv(".. /input/battles.csv")
pokemon = pd.read_csv(".. /input/pokemon.csv")
pokemon.describe()<feature_engineering> | app_train['TARGET'].value_counts()
print('The proportion of label 1 is %.2f' %(sum(app_train['TARGET']==1)/app_train.shape[0]*100), '%' ) | Home Credit Default Risk |
1,316,642 | def normalizeColumns(columns, dataset):
for column in columns:
dataset[column] =(dataset[column] - dataset[column].mean())/ dataset[column].std()<categorify> | def missing_values_table(df):
mis_val = df.isnull().sum()
mis_val_percent = df.isnull().sum() * 100 / df.shape[0]
mis_val_table = pd.concat([mis_val, mis_val_percent], axis = 1)
mis_val_table_rename_columns = mis_val_table.rename(columns = {0: 'Missing Values', 1: 'Percentage'})
mis_val_table_rename_columns = mis_val... | Home Credit Default Risk |
1,316,642 | pokemon.fillna("noType", inplace=True)
pokemon = pd.concat([
pokemon,
pd.get_dummies(pokemon["Type 1"], prefix="t1"),
pd.get_dummies(pokemon["Type 2"], prefix="t2"),
], axis=1)
pokemon.drop(["Type 1", "Type 2", "Name", "Generation", "Legendary"], axis=1, inplace=True)
normalizeColumns(["HP", "Sp.Atk", "Sp.Def", "Att... | missing_values = missing_values_table(app_train)
missing_values.head(20 ) | Home Credit Default Risk |
1,316,642 | def mergePokemonStats(battles, pokemon):
data = battles \
.merge(pokemon, left_on="First_pokemon", right_on="
.merge(pokemon, left_on="Second_pokemon", right_on="
.sort_values(['battle_number'])\
.drop(["
data = data.reindex(sorted(data.columns), axis=1)
return data<merge> | app_train.dtypes.value_counts() | Home Credit Default Risk |
1,316,642 | data = mergePokemonStats(battles, pokemon)
data.head()<prepare_x_and_y> | app_train.select_dtypes('object' ).apply(pd.Series.nunique, axis=0 ) | Home Credit Default Risk |
1,316,642 | y = data["Winner"].values
x = data.drop(["Winner"],axis=1 ).values<import_modules> | app_test.dtypes.value_counts() | Home Credit Default Risk |
1,316,642 | import tensorflow as tf
import sklearn.model_selection
import sklearn.ensemble
import sklearn.metrics<choose_model_class> | app_train = pd.get_dummies(app_train)
app_test = pd.get_dummies(app_test)
print('Training Features shape: ', app_train.shape)
print('Testing Features shape: ', app_test.shape ) | Home Credit Default Risk |
1,316,642 | mlp = tf.keras.Sequential([
tf.keras.layers.Dense(310, activation=tf.nn.relu),
tf.keras.layers.GaussianNoise(0.17),
tf.keras.layers.Dropout(0.50),
tf.keras.layers.Dense(310, activation=tf.nn.relu),
tf.keras.layers.Dropout(0.50),
tf.keras.layers.Dense(1, activation=tf.nn.sigmoid)
] )<choose_model_class> | train_labels = app_train['TARGET']
app_train, app_test = app_train.align(app_test, join = 'inner', axis = 1)
app_train['TARGET'] = train_labels
print('Training Features shape: ', app_train.shape)
print('Testing Features shape: ', app_test.shape ) | Home Credit Default Risk |
1,316,642 | mlp.compile(optimizer="adam",
loss="binary_crossentropy",
metrics=["accuracy"] )<train_model> | app_train['DAYS_EMPLOYED_ANOM'] = app_train['DAYS_EMPLOYED'] == 365243
app_train['DAYS_EMPLOYED'].replace({365243: np.nan}, inplace = True)
app_test['DAYS_EMPLOYED_ANOM'] = app_test['DAYS_EMPLOYED'] == 365243
app_test['DAYS_EMPLOYED'].replace({365243: np.nan}, inplace = True)
print('Training Features shape: ', app_tr... | Home Credit Default Risk |
1,316,642 | history = mlp.fit(
x,
y,
epochs=350,
workers=4,
batch_size=512,
shuffle=True,
validation_split=0.1
)<split> | app_train_domain = app_train.copy()
app_test_domain = app_test.copy()
app_train_domain['CREDIT_INCOME_PERCENT'] = app_train_domain['AMT_CREDIT'] / app_train_domain['AMT_INCOME_TOTAL']
app_train_domain['ANNUITY_INCOME_PERCENT'] = app_train_domain['AMT_ANNUITY'] / app_train_domain['AMT_INCOME_TOTAL']
app_train_domain['CR... | Home Credit Default Risk |
1,316,642 | x_train, x_test, y_train, y_test = sklearn.model_selection.train_test_split(x, y, test_size=0.1)
gradientBoosting = sklearn.ensemble.GradientBoostingClassifier(
n_estimators=1000,
max_depth=10,
verbose = 1
)
gradientBoosting.fit(x_train, y_train )<compute_test_metric> | bureau = pd.read_csv('.. /input/bureau.csv')
bureau.head() | Home Credit Default Risk |
1,316,642 | sklearn.metrics.accuracy_score(y_test, gradientBoosting.predict(x_test), normalize=True, sample_weight=None )<load_from_csv> | previous_loan_counts = bureau.groupby('SK_ID_CURR', as_index=False)['SK_ID_BUREAU'].count().rename(columns = {'SK_ID_BUREAU': 'previous_loan_counts'})
previous_loan_counts.head()
| Home Credit Default Risk |
1,316,642 | submission = pd.read_csv(".. /input/test.csv")
test = mergePokemonStats(submission, pokemon )<predict_on_test> | def agg_numeric(df, group_var, df_name):
for col in df:
if col != group_var and 'SK_ID' in col:
df = df.drop(columns = col)
group_ids = df[group_var]
numeric_df = df.select_dtypes('number')
numeric_df[group_var] = group_ids
agg = numeric_df.groupby(group_var ).agg(['count', 'mean', 'max', 'min', 'sum'] ).reset_inde... | Home Credit Default Risk |
1,316,642 | predictionMLP = np.transpose(mlp.predict(test.values))
predictionGB = gradientBoosting.predict_proba(test.values)
predictionGB = np.array([predict[1] for predict in predictionGB])
prediction =(( predictionMLP)+(0.10*predictionGB)) /1.10
prediction[prediction <= 0.5] = 0
prediction[prediction > 0.5] = 1<data_type_conv... | def count_categorical(df, group_var, df_name):
categorical = pd.get_dummies(df.select_dtypes('object'))
categorical[group_var] = df[group_var]
categorical = categorical.groupby(group_var ).agg(['sum', 'mean'])
column_names = []
for var in categorical.columns.levels[0]:
for stat in ['count', 'count_norm']:
column_nam... | Home Credit Default Risk |
1,316,642 | submission["Winner"] = prediction[0].astype("int")
submission.drop(["First_pokemon", "Second_pokemon"], axis=1, inplace=True)
submission.head()<save_to_csv> | train = app_train_domain.merge(previous_loan_counts, on = 'SK_ID_CURR', how = 'left')
train['previous_loan_counts'] = train['previous_loan_counts'].fillna(0)
test = app_test_domain.merge(previous_loan_counts, on = 'SK_ID_CURR', how = 'left')
test['previous_loan_counts'] = test['previous_loan_counts'].fillna(0 ) | Home Credit Default Risk |
1,316,642 | submission.to_csv("submission.csv", index=False )<save_to_csv> | train = train.merge(bureau_agg, on = 'SK_ID_CURR', how = 'left')
test = test.merge(bureau_agg, on = 'SK_ID_CURR', how = 'left' ) | Home Credit Default Risk |
1,316,642 | submission.to_csv("submission.csv", index=False )<load_from_csv> | train = train.merge(bureau_counts, on = 'SK_ID_CURR', how = 'left')
test = test.merge(bureau_counts, on = 'SK_ID_CURR', how = 'left' ) | Home Credit Default Risk |
1,316,642 | battles = pd.read_csv('.. /input/battles.csv')
pokemon = pd.read_csv('.. /input/pokemon.csv')
test = pd.read_csv('.. /input/test.csv')
pokemon = pokemon.drop(columns='Name', errors='ignore')
pokemon.loc[:, 'Type 1'] = pokemon.loc[:, 'Type 1'].fillna('None')
pokemon.loc[:, 'Type 2'] = pokemon.loc[:, 'Type 2'].filln... | print('Before align train.shape: ', train.shape)
print('Before align test.shape: ', test.shape)
train_labels = train['TARGET']
train, test = train.align(test, join = 'inner', axis = 1)
train['TARGET'] = train_labels
print('After align train.shape: ', train.shape)
print('After align test.shape: ', test.shape ) | Home Credit Default Risk |
1,316,642 | tout = pd.merge(test, pokemon, left_on='First_pokemon' ,right_on='
tout = pd.merge(tout, pokemon, left_on='Second_pokemon',right_on='
tout = tout.sort_values(['battle_number'])
tout = tout.drop(columns=['
tout = tout.iloc[:, 3:]
tout.head()<split> | bureau_balance = pd.read_csv('.. /input/bureau_balance.csv')
bureau_balance.head() | Home Credit Default Risk |
1,316,642 | TEST_SIZE = 0.01
tr_idx, ts_idx = train_test_split(range(len(battles)) , test_size=TEST_SIZE)
cat_names = ['Generation_A', 'Generation_B', 'Type 1_A', 'Type 1_B', 'Type 2_A', 'Type 2_B', 'Legendary_A', 'Legendary_B']
procs = [Categorify, Normalize]
db = TabularDataBunch.from_df(path='.', df=data, dep_var='Winner', val... | bureau_balance_agg = agg_numeric(bureau_balance, group_var = 'SK_ID_BUREAU', df_name = 'bureau_balance')
bureau_balance_agg.head() | Home Credit Default Risk |
1,316,642 | learn.fit_one_cycle(50, 1e-1 )<prepare_output> | bureau_balance_counts = count_categorical(bureau_balance, group_var = 'SK_ID_BUREAU', df_name = 'bureau_balance')
bureau_balance_counts.head() | Home Credit Default Risk |
1,316,642 | preds, _ = learn.get_preds(ds_type=DatasetType.Test)
preds = np.argmax(preds, axis=1 ).numpy()
submission = pd.DataFrame(test.iloc[:, 0])
submission['Winner'] = preds<save_to_csv> | bureau_by_loan = bureau_balance_agg.merge(bureau_balance_counts, right_index = True, left_on = 'SK_ID_BUREAU', how = 'outer' ) | Home Credit Default Risk |
1,316,642 | print(submission.head())
submission.to_csv('./submission.csv', index=False )<save_to_csv> | bureau_by_loan = bureau_by_loan.merge(bureau[['SK_ID_BUREAU', 'SK_ID_CURR']], on = 'SK_ID_BUREAU', how = 'left' ) | Home Credit Default Risk |
1,316,642 | def create_download_link(df, title = "Download CSV file", filename = "data.csv"):
csv = df.to_csv(index=False)
b64 = base64.b64encode(csv.encode())
payload = b64.decode()
html = '<a download="{filename}" href="data:text/csv;base64,{payload}" target="_blank">{title}</a>'
html = html.format(payload=payload,title=title,... | bureau_balance_by_client = agg_numeric(bureau_by_loan.drop(columns = ['SK_ID_BUREAU']), group_var = 'SK_ID_CURR', df_name = 'client')
bureau_balance_by_client.head() | Home Credit Default Risk |
1,316,642 | %matplotlib inline<load_from_csv> | train = train.merge(bureau_balance_by_client, on = 'SK_ID_CURR', how = 'left')
test = test.merge(bureau_balance_by_client, on = 'SK_ID_CURR', how = 'left' ) | Home Credit Default Risk |
1,316,642 | train = pd.read_csv('.. /input/train.csv')
test = pd.read_csv('.. /input/test.csv' )<load_from_csv> | print('Before align train.shape: ', train.shape)
print('Before align test.shape: ', test.shape)
train_labels = train['TARGET']
train, test = train.align(test, join = 'inner', axis = 1)
train['TARGET'] = train_labels
print('After align train.shape: ', train.shape)
print('After align test.shape: ', test.shape ) | Home Credit Default Risk |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.