pandas 分组聚合

import pandas as pdimport numpy as npdf = pd.DataFrame({'key1': ['a', 'a', 'b', 'b', 'a'],                'key2': ['one', 'two', 'one', 'two', 'one'],                'data1': np.random.randn(5),                'data2': np.random.randn(5)})df

data1 data2 key1 key2 0 1.244597 1.243988 a one 1 -1.253387 -0.544606 a two 2 -1.273505 1.017159 b one 3 -0.681958 -0.187279 b two 4 -0.683225 1.054060 a one

---

按‘key1’分组

grouped = df.groupby('key1')grouped#<pandas.core.groupby.DataFrameGroupBy object at 0x0000000009102908>

---

对于分组的某一列或者多个列，应用agg(func)可以对分组后的数据应用func函数。例如：用group1[‘data1’].agg(‘mean’)对分组后的’data1’列求均值。当然也可以推广到同时作用于多个列和使用多个函数上。

grouped.agg('mean')

data1 data2 key1 a -0.230672 0.584481 b -0.977731 0.414940

grouped.agg(['sum','mean'])

data1 data2 sum mean sum mean key1 a -0.692015 -0.230672 1.753442 0.584481 b -1.955463 -0.977731 0.829881 0.414940

grouped['data1'].agg(['sum','mean'])

sum mean key1 a -0.692015 -0.230672 b -1.955463 -0.977731

grouped[['data1']].agg(['sum','mean'])

data1 sum mean key1 a -0.692015 -0.230672 b -1.955463 -0.977731

pd.pivot_table(df, index='key1')

data1 data2 key1 a -0.230672 0.584481 b -0.977731 0.414940

pd.pivot_table(df, index='key1', columns='key2') 

data1 data2 key2 one two one two key1 a 0.280686 -1.253387 1.149024 -0.544606 b -1.273505 -0.681958 1.017159 -0.187279

---

透视表pivot_table，单元格中的数值为同组数值的平均值

df.pivot_table(['data1'], index='key1',columns='key2')  

data1 key2 one two key1 a 0.280686 -1.253387 b -1.273505 -0.681958

df

data1 data2 key1 key2 0 1.244597 1.243988 a one 1 -1.253387 -0.544606 a two 2 -1.273505 1.017159 b one 3 -0.681958 -0.187279 b two 4 -0.683225 1.054060 a one

如果将参数margins设置为True，则可以得到分项总计数据。

df.pivot_table(index='key1',columns='key2', margins=True)

data1 data2 key2 one two All one two All key1 a 0.280686 -1.253387 -0.230672 1.149024 -0.544606 0.584481 b -1.273505 -0.681958 -0.977731 1.017159 -0.187279 0.414940 All -0.237378 -0.967672 -0.529496 1.105069 -0.365942 0.516665

---

按’key1’,’key2’进行分组，使用describe()列出详细信息

grouped2=df.groupby(['key1','key2'])grouped2.describe()

data1 data2 count mean std min 25% 50% 75% max count mean std min 25% 50% 75% max key1 key2 a one 2.0 0.280686 1.363176 -0.683225 -0.201270 0.280686 0.762641 1.244597 2.0 1.149024 0.134299 1.054060 1.101542 1.149024 1.196506 1.243988 two 1.0 -1.253387 NaN -1.253387 -1.253387 -1.253387 -1.253387 -1.253387 1.0 -0.544606 NaN -0.544606 -0.544606 -0.544606 -0.544606 -0.544606 b one 1.0 -1.273505 NaN -1.273505 -1.273505 -1.273505 -1.273505 -1.273505 1.0 1.017159 NaN 1.017159 1.017159 1.017159 1.017159 1.017159 two 1.0 -0.681958 NaN -0.681958 -0.681958 -0.681958 -0.681958 -0.681958 1.0 -0.187279 NaN -0.187279 -0.187279 -0.187279 -0.187279 -0.187279