Python DataFrame.groupby()聚合函数,分组级运算

import pandas as pd
df = pd.DataFrame({'key1':list('ababa'),
  'key2': ['one','two','one','two','one'],
  'data1': np.random.randn(5),
  'data2': np.random.randn(5)})
print(df)

 data1 data2 key1 key2
0 -1.313101 -0.453361 a one
1 0.791463 1.096693 b two
2 0.462611 1.150597 a one
3 -0.216121 1.381333 b two
4 0.077367 -0.282876 a one

#将df['data1']按照分组键为df['key1']进行分组
grouped=df['data1'].groupby(df['key1'])
print(grouped.mean())
key1
a -0.257707
b 0.287671
Name: data1, dtype: float64
states=np.array(['Ohio','California','California','Ohio','Ohio'])
years=np.array([2005,2005,2006,2005,2006])
#states第一层索引，years第二层分层索引
print(df['data1'].groupby([states,years]).mean())
California 2005 0.791463
 2006 0.462611
Ohio 2005 -0.764611
 2006 0.077367
Name: data1, dtype: float64
#df根据‘key1'分组，然后对df剩余数值型的数据运算
df.groupby('key1').mean()
 data1 data2
key1  
a -0.257707 0.138120
b 0.287671 1.239013
#可以看出没有key2列，因为df[‘key2']不是数值数据，所以被从结果中移除。默认情况下，所有数值列都会被聚合，虽然有时可能被过滤为一个子集。

#name就是groupby中的key1的值，group就是要输出的内容
for name, group in df.groupby('key1'):
 print (name,group)
a data1 data2 key1 key2
0 -1.313101 -0.453361 a one
2 0.462611 1.150597 a one
4 0.077367 -0.282876 a one
b data1 data2 key1 key2
1 0.791463 1.096693 b two
3 -0.216121 1.381333 b two

#转化为字典
piece=dict(list(df.groupby('key1')))
{'a': data1 data2 key1 key2
 0 -1.313101 -0.453361 a one
 2 0.462611 1.150597 a one
 4 0.077367 -0.282876 a one, 'b': data1 data2 key1 key2
 1 0.791463 1.096693 b two
 3 -0.216121 1.381333 b two}
#对字典取值
value = piece['a']

grouped=df.groupby(df.dtypes, axis=1)
value = dict(list(grouped))
print(value)
{dtype('float64'): data1 data2
0 -1.313101 -0.453361
1 0.791463 1.096693
2 0.462611 1.150597
3 -0.216121 1.381333
4 0.077367 -0.282876, dtype('O'): key1 key2
0 a one
1 b two
2 a one
3 b two
4 a one}

#对df进行'key1'，'key2'的两次分组，然后取data2的数据，对两次细分的分组数据取均值
value = df.groupby(['key1','key2'])[['data2']].mean()
 data2
key1 key2 
a one 0.138120
b two 1.239013
----------------------------------
df
Out[1]: 
 data1 data2 key1 key2
0 -1.313101 -0.453361 a one
1 0.791463 1.096693 b two
2 0.462611 1.150597 a one
3 -0.216121 1.381333 b two
4 0.077367 -0.282876 a one
----------------------------------
df['key2'].iloc[-1] ='two'
value = df.groupby(['key1','key2'])[['data2']].mean()
value
Out[2]: 
 data2
key1 key2 
a one 0.348618
 two -0.282876
b two 1.239013

from operator import itemgetter #itemgetter用来去dict中的key，省去了使用lambda函数
from itertools import groupby #itertool还包含有其他很多函数，比如将多个list联合起来。。
d1={'name':'zhangsan','age':20,'country':'China'}
d2={'name':'wangwu','age':19,'country':'USA'}
d3={'name':'lisi','age':22,'country':'JP'}
d4={'name':'zhaoliu','age':22,'country':'USA'}
d5={'name':'pengqi','age':22,'country':'USA'}
d6={'name':'lijiu','age':22,'country':'China'}
lst=[d1,d2,d3,d4,d5,d6]
#通过country进行分组：
lst.sort(key=itemgetter('country')) #需要先排序，然后才能groupby。lst排序后自身被改变
lstg = groupby(lst,itemgetter('country')) 
#lstg = groupby(lst,key=lambda x:x['country']) 等同于使用itemgetter()
for key,group in lstg:
 for g in group: #group是一个迭代器，包含了所有的分组列表
 print key,g
返回：
China {'country': 'China', 'age': 20, 'name': 'zhangsan'}
China {'country': 'China', 'age': 22, 'name': 'lijiu'}
JP {'country': 'JP', 'age': 22, 'name': 'lisi'}
USA {'country': 'USA', 'age': 19, 'name': 'wangwu'}
USA {'country': 'USA', 'age': 22, 'name': 'zhaoliu'}
USA {'country': 'USA', 'age': 22, 'name': 'pengqi'}
print [key for key,group in lstg] #返回：['China', 'JP', 'USA']
print [(key,list(group)) for key,group in lstg]
#返回的list中包含着三个元组：
[('China', [{'country': 'China', 'age': 20, 'name': 'zhangsan'}, {'country': 'China', 'age': 22, 'name': 'lijiu'}]), ('JP', [{'country': 'JP', 'age': 22, 'name': 'lisi'}]), ('USA', [{'country': 'USA', 'age': 19, 'name': 'wangwu'}, {'country': 'USA', 'age': 22, 'name': 'zhaoliu'}, {'country': 'USA', 'age': 22, 'name': 'pengqi'}])]
print dict([(key,list(group)) for key,group in lstg])
#返回的是一个字典：
{'JP': [{'country': 'JP', 'age': 22, 'name': 'lisi'}], 'China': [{'country': 'China', 'age': 20, 'name': 'zhangsan'}, {'country': 'China', 'age': 22, 'name': 'lijiu'}], 'USA': [{'country': 'USA', 'age': 19, 'name': 'wangwu'}, {'country': 'USA', 'age': 22, 'name': 'zhaoliu'}, {'country': 'USA', 'age': 22, 'name': 'pengqi'}]}
print dict([(key,len(list(group))) for key,group in lstg])
#返回每个分组的个数：
{'JP': 1, 'China': 2, 'USA': 3}

#返回包含有2个以上元素的分组
print [key for key,group in groupby(sorted(lst,key=itemgetter('country')),itemgetter('country')) if len(list(group))>=2]
#返回：['China', 'USA']

lstg = groupby(sorted(lst,key=itemgetter('country')),key=itemgetter('country')) 
lstgall=[(key,list(group)) for key,group in lstg ]
print dict(filter(lambda x:len(x[1])>2,lstgall)) 
#过滤出分组后的元素个数大于2个的分组，返回：
{'USA': [{'country': 'USA', 'age': 19, 'name': 'wangwu'}, {'country': 'USA', 'age': 22, 'name': 'zhaoliu'}, {'country': 'USA', 'age': 22, 'name': 'pengqi'}]}

自定义分组：

from itertools import groupby
lst=[2,8,11,25,43,6,9,29,51,66]

def gb(num):
 if num <= 10:
 return 'less'
 elif num >=30:
 return 'great'
 else:
 return 'middle'
print [(k,list(g))for k,g in groupby(sorted(lst),key=gb)]
返回：
[('less', [2, 6, 8, 9]), ('middle', [11, 25, 29]), ('great', [43, 51, 66])]