def last_k_instalment_features_with_fractions(gr, periods, fraction_periods):
gr_ = gr.copy()
features = {}
features_temp = {}
for period in periods:
gr_period = gr_[gr_['days'] <= period]
features_temp = utils.add_features_in_group(
features_temp, gr_period, 'installments',
['mean', 'var', 'skew', 'kurt', 'iqr'], 'last_{}_'.format(period))
features_temp = utils.add_features_in_group(
features_temp, gr_period, 'purchase_amount',
['sum', 'max', 'mean', 'var', 'skew', 'kurt', 'iqr'],
'last_{}_'.format(period))
for short_period, long_period in fraction_periods:
short_feature_names = utils._get_feature_names(features_temp,
short_period)
long_feature_names = utils._get_feature_names(features_temp,
long_period)
for short_feature, long_feature in zip(short_feature_names,
long_feature_names):
old_name_chunk = '_{}_'.format(short_period)
new_name_chunk = '_{}by{}_fraction_'.format(
short_period, long_period)
fraction_feature_name = short_feature.replace(
old_name_chunk, new_name_chunk)
features[fraction_feature_name] = utils.safe_div(
features_temp[short_feature], features_temp[long_feature])
return pd.Series(features)
def _installments_last_loan_features(gr):
"""Return statistics for the last loan.
Returns the sum, mean, max and std for the
last loan in the given group, which is usually
grouped by customer id.
Arguments:
gr: Pandas groupby object.
Returns:
features: dictionary with features.
"""
gr_ = gr.copy()
gr_.sort_values(['DAYS_INSTALMENT'], ascending=False, inplace=True)
last_installment_id = gr_['SK_ID_PREV'].iloc[0]
gr_ = gr_[gr_['SK_ID_PREV'] == last_installment_id]
features = {}
features = utils.add_features_in_group(features, gr_, 'DPD',
['sum', 'mean', 'max', 'std'],
'LAST_LOAN_')
features = utils.add_features_in_group(features, gr_, 'LATE_PAYMENT',
['count', 'mean'],
'LAST_LOAN_')
features = utils.add_features_in_group(features, gr_, 'PAID_OVER_AMOUNT',
['sum', 'mean', 'max', 'min', 'std'],
'LAST_LOAN_')
features = utils.add_features_in_group(features, gr_, 'PAID_OVER',
['count', 'mean'],
'LAST_LOAN_')
return features
def last_k_instalment_features(gr, periods):
gr_ = gr.copy()
features = {}
for period in periods:
gr_period = gr_[gr_['days'] <= period]
features = utils.add_features_in_group(features, gr_period,
'installments',
['max', 'mean', 'var', 'skew'],
'hist_last_{}_'.format(period))
features = utils.add_features_in_group(
features, gr_period, 'purchase_amount',
['sum', 'max', 'min', 'mean', 'var', 'skew'],
'hist_last_{}_'.format(period))
return features
def last_link_time_features(gr,periods):
gr_ = gr.copy()
gr_ = gr_.iloc[::-1]
features = {}
for period in periods:
if period > 10e5:
period_name = 'zsl_all_'
gr_period = gr_.copy()
else:
period_name = 'zsl_last_{}_'.format(period)
gr_period = gr_.iloc[:period]
features = add_features_in_group(features, gr_period, 'link_time',
['max', 'sum', 'mean','min','skew','std'],
period_name)
features = add_features_in_group(features, gr_period, 'link_current_status',
['mean', 'nunique'],
period_name)
return features
def last_k_link_time_interval(gr, periods):
gr_ = gr.copy()
gr_ = gr_.iloc[::-1]
gr_['t_i_v'] = gr_['link_time'].diff()
gr_['t_i_v'] = gr_['t_i_v']
gr_['t_i_v'] = gr_['t_i_v'].fillna(0)
gr_['c_s_v'] = gr_['link_current_status'].diff()
gr_['c_s_v'] = gr_['c_s_v']
gr_['c_s_v'] = gr_['c_s_v'].fillna(0)
gr_ = gr_.drop_duplicates().reset_index(drop = True)
# link time变化
features = {}
for period in periods:
if period > 10e5:
period_name = 'zsl_link_time_interval_all'
gr_period = gr_.copy()
else:
period_name = 'zsl_link_time_interval_last_{}_'.format(period)
gr_period = gr_.iloc[:period]
features = add_features_in_group(features, gr_period, 't_i_v',
['mean','max', 'min', 'std','skew','sum'],
# ['diff'],
period_name)
# current status变化
for period in periods:
if period > 10e5:
period_name = 'zsl_link_current_status_interval_all'
gr_period = gr_.copy()
else:
period_name = 'zsl_link_current_status_interval_last_{}_'.format(period)
gr_period = gr_.iloc[:period]
features = add_features_in_group(features, gr_period, 'c_s_v',
['mean', 'std', 'skew'],
# ['diff'],
period_name)
return features