conditional_features.py

conditional_features.py

import pandas as pd


def build_conditional_features(df, cond_series, feat_cols=None):
    """
    return a new dataframe with conditional features
    """
    if len(cond_series) != len(df):
        raise Exception("Condition series isn't same num rows as features")
    if feat_cols is None:
        feat_cols = list(df.columns)
    cond_feat_to_series = {}
    for target_cond in cond_series.unique():
        cond_mask = cond_series.map(lambda c: int(c == target_cond))
        for feat in feat_cols:
            cond_feat = f"{feat}_||_{target_cond}"
            cond_feat_to_series[cond_feat] = df[feat] * cond_mask
    return pd.DataFrame(data=cond_feat_to_series)


raw_df = pd.DataFrame(data={'name': ['foo','bar','roo'], 'age': [32, 20, 35]})
feats_df = pd.DataFrame([[1,1,1],[0,1,1],[1,0,0]], columns=['f1', 'f2', 'f3'])


>>> feats_df
   f1  f2  f3
0   1   1   1
1   0   1   1
2   1   0   0



def age_condition(row):
    if row['age'] > 21:
        return 'adult'
    return 'minor'


cond_series = raw_df.apply(age_condition, axis=1)
cond_feats = build_conditional_features(feats_df, cond_series)


>>> cond_feats
   f1_||_adult  f2_||_adult  f3_||_adult  f1_||_minor  f2_||_minor  f3_||_minor
0            1            1            1            0            0            0
1            0            0            0            0            1            1
2            1            0            0            0            0            0


feats_df = pd.concat([feats_df, cond_feats], axis=1)