series中MultiIndex

1.dataFrame中groupby後變成MultiIndex型的series

group=df_tmp.groupby(['SK_ID_CURR','CREDIT_ACTIVE']).size()
group      

SK_ID_CURR  CREDIT_ACTIVE
162297      Active           1
            Closed           2
215354      Active           6
            Closed           1
dtype: int64      

group.index      

MultiIndex(levels=[[162297, 215354], ['Active', 'Closed']],
           labels=[[0, 0, 1, 1], [0, 1, 0, 1]],
           names=['SK_ID_CURR', 'CREDIT_ACTIVE'])      

for id in group.index.levels[0]:
    print (id)      

162297
215354      

#group.index.get_loc((215354,"Active"))擷取索引為(215354,"Active")其對應的values中的下标      

group.values[group.index.get_loc((215354,"Active"))]      

6

2.衍生出各個id，Active的數量，代碼如下：      

df_new=pd.DataFrame({'SK_ID_CURR':list(set(df_bureau.SK_ID_CURR))}) #衍生特征首先建構一個新的dataframe      

group=df_bureau.groupby(['SK_ID_CURR','CREDIT_ACTIVE']).size()

df_new['CREDIT_ACTIVE_COUNT']=0
for id in group.index.levels[0]:
    df_new['CREDIT_ACTIVE_COUNT'].loc[df_new.SK_ID_CURR==id]=group.values[group.index.get_loc((id,"Active"))]