python group_by

-1.类似字典的调用方式

groupby = df.groupby('id')
# 遍历
for id,group in groupby:
    print(id)
    print(group)
# 取其中一组
data = groupby.get_group('aaa')

-2.单列函数

df_groupby = df.groupby('id')['salary'].sum()
df_groupby = df.groupby(['id','gender'])['salary'].sum()

mean, min, max, count, std,
https://www.cnblogs.com/Yanjy-OnlyOne/p/11217802.html

-3.agg

df_groupby = df.groupby('id')['salary'].agg(['max','min','std'])

-4.自定义函数

效果：各类别高于3倍标准差为1，否则为0

    def handle_abnormal(df, shred=3):
        df_temp = df[df>(df.mean()+ shred *df.std())]
        if not df_temp.empty:
            return 1
        else:
            return 0
        
    df_beh_abnormal = df_beh_date.groupby('id')['count'].apply(handle_abnormal)