Pandas基础二重点 -----------分组聚合、agg函数、apply函数（可传参）

一、分组 groupby

1.1 如何对二维数组分组查询

import pandas as pd
import numpy as np
df1=pd.DataFrame({
    'name':['lilei','rose','jack','haha','rose','jack'],
    'expend':[100,200,300,500,800,100],
    'year':[2019,2018,2019,2019,2017,2017],
    'salary':[1000,2000,3000,3000,4000,500]    
})
print(df1)
group_name=df1.groupby('name')
#分组后是对象
print(group_name,type(group_name))
#查看分组概览
print(group_name.groups)
#分组后数量统计
print(group_name.count())

name expend year salary
0 lilei 100 2019 1000
1 rose 200 2018 2000
2 jack 300 2019 3000
3 haha 500 2019 3000
4 rose 800 2017 4000
5 jack 100 2017 500
#分组后是对象
<pandas.core.groupby.groupby.DataFrameGroupBy object at 0x000002C106149518> <class ‘pandas.core.groupby.groupby.DataFrameGroupBy’>
#查看分组概览
{‘haha’: Int64Index([3], dtype=‘int64’), ‘jack’: Int64Index([2, 5], dtype=‘int64’), ‘lilei’: Int64Index([0], dtype=‘int64’), ‘rose’: Int64Index([1, 4], dtype=‘int64’)}
#分组后数量统计
expend year salary
name
haha 1 1 1
jack 2 2 2
lilei 1 1 1
rose 2 2 2

#遍历分组后的数据
for name,group in group_name:
    display(group)

#按多列进行分组
group_two_name=df1.groupby(['name','year'])
print(group_two_name.groups)

1.2 使用pd.cut分段

1. 先使用pd.cut对目标数据分段
2. 按照分段后的数据进行分组
3. 可使用交叉表

import pandas as pd
from pandas import Series,DataFrame
import numpy as np
import random
df2=pd.DataFrame({
    'age':[20,31,41,66,67],
    'sex':['M','M','M','F','F']
})
print(df2)
print("================================================")
#将数据按段分成不同范围
cut_age=pd.cut(df2['age'],bins=[19,40,65,100])
print(cut_age)
print("================================================")
#按照上述范围对数据分组
group_age=df2.groupby(cut_age).count()
print(group_age)
print("================================================")
#交叉表，统计分组后，‘sex’出现的频次
print(pd.crosstab(cut_age,df2['sex'],margins=True))

age sex
0 20 M
1 31 M
2 41 M
3 66 F
4 67 F
================================================
0 (19, 40]
1 (19, 40]
2 (40, 65]
3 (65, 100]
4 (65, 100]
Name: age, dtype: category
Categories (3, interval[int64]): [(19, 40] < (40, 65] < (65, 100]]
================================================
age sex
age
(19, 40] 2 2
(40, 65] 1 1
(65, 100] 2 2
================================================
sex F M All
age
(19, 40] 0 2 2
(40, 65] 0 1 1
(65, 100] 2 0 2
All 2 3 5

二、聚合

2.1 分组后直接聚合

分组后直接求和

取某列的结果：

• 可先取出对应列，后分组
• 也可先分组后取对应列

df3=pd.DataFrame({
    'data1':np.random.randint(1,10,5),
    'data2':np.random.randint(11,20,5),
    'key1':list('aabba'),
    'key2':list('xyyxy')
})
print(df3)
print("================================================")
#分组后直接聚合计算，不是数值的列（麻烦列）会被清除
print(df3.groupby('key1').sum())
print("================================================")
#只想要data1的数据，那么可以先取出对应列，后分组
print(df3['data1'].groupby(df3['key1']).sum())
#也可以先分组后取对应列。效率基本无差别
print(df3.groupby('key1')['data1'].sum())

data1 data2 key1 key2
0 1 17 a x
1 1 15 a y
2 7 18 b y
3 1 11 b x
4 8 11 a y
================================================
data1 data2
key1
a 10 43
b 8 29
================================================
key1
a 10
b 8
Name: data1, dtype: int32
key1
a 10
b 8
Name: data1, dtype: int32
1
​

2.2 agg（）函数，可同时做多个聚合

print(df3.groupby('key1').agg('sum'))
print("================================================")
display(df3.groupby('key1').agg(['sum','mean','std']))

2.3 ※※自定义函数，传入agg方法中使用

日常可自定义聚合运算，传入agg（）中调用
agg中使用元祖可给自定义函数起别名
注：只有pandas中函数作为参数传给agg，故函数不用加括号

import pandas as pd
from pandas import Series,DataFrame
import numpy as np
import random
df4=pd.DataFrame({
    'data1':np.random.randint(1,10,5),
    'data2':np.random.randint(11,20,5),
    'key1':list('aabba'),
    'key2':list('xyyxy')
})
print(df4)
def num_range(df):
    """
    返回最大值与最小值的差值
    """
    return df.max() - df.min()
print(df4.groupby('key1').agg(num_range))
#重命名自定义函数
print(df4.groupby('key1').agg(['sum',('range',num_range)]))
#agg中传入字典，key代表对应列，value代表对应列的聚合函数
data_map={
    'data1':['mean','max'],
    'data2':'sum'
}
print(df4.groupby('key1').agg(data_map))

2.4 apply函数，自由度最高的，可传参

跟agg函数相比，apply可给函数添加参数

DataFrame.apply(func, axis=0, broadcast=False, raw=False, reduce=None, args=(), **kwds)

df1=pd.DataFrame({'sex':list('FFMFMMF'),'smoker':list('YNYYNYY'),'age':[21,30,17,37,40,18,26],'weight':[120,100,132,140,94,89,123]})
print(df1)
def bin_age(age):
    if age>=18:
        return 1
    else:
        return 0
#改变age列的值
print(df1['age'].apply(bin_age))
df1['age']=df1['age'].apply(bin_age)
print(df1)
#分别取出抽烟和不抽烟的体重前2
def weight_top(self,col,n=5):
    return self.sort_values(by=col)[-n:]
df1.groupby('smoker').apply(weight_top,col='weight',n=2)
#另一种写法
#df1.groupby('smoker').apply(weight_top,args=('weight',2))