def install_trend_k_feature(df, origin):
func = partial(trend_in_last_k_instalment_features, periods=[5, 10, 50])
groupby = origin.groupby(['SK_ID_CURR'])
g = parallel_apply(groupby, func, index_name='SK_ID_CURR',
num_workers=4).reset_index()
df = pd.merge(df, g, on=['SK_ID_CURR'], how='left')
return df
def pos_cash_trend_installment(df, origin):
groupby = origin.groupby(['SK_ID_CURR'])
func = partial(trend_in_last_k_installment_features,
periods=[1, 6, 12, 30, 60])
g = parallel_apply(groupby, func, index_name='SK_ID_CURR',
num_workers=8).reset_index()
df = pd.merge(df, g, on=['SK_ID_CURR'], how='left')
return df
def pos_cash_last_loan(df, origin):
groupby = origin.groupby(['SK_ID_CURR'])
g = parallel_apply(groupby,
last_loan_features,
index_name='SK_ID_CURR',
num_workers=8).reset_index()
df = pd.merge(df, g, on=['SK_ID_CURR'], how='left')
return df
def install_last_k_feature(df, origin):
func = partial(last_k_instalment_features, periods=[1, 5, 10, 20, 50, 100])
groupby = origin.groupby(['SK_ID_CURR'])
g = parallel_apply(groupby, func, index_name='SK_ID_CURR',
num_workers=8).reset_index()
g = correlation_reduce(g)
df = pd.merge(df, g, on=['SK_ID_CURR'], how='left')
return df
def bureau_trend_features(df, bureau):
groupby = bureau.groupby(['SK_ID_CURR'])
g = parallel_apply(groupby, trend_in_last_k_bureau_features,
index_name = 'SK_ID_CURR', num_workers = 8).reset_index()
df = pd.merge(df, g, on = ['SK_ID_CURR'], validate = 'one_to_one')
return df