pandas groupby教程

pandas groupby

按功能区分

• 分组数据

  df = pd.DataFrame({'Year' : ['2001', '2002', '2001', '2002',
                            '2001', '2002', '2001', '2002'],
                     'score' : ['primary', 'second', 'third', 'fourth',
                            'primary', 'second', 'fourth', 'third'],
                     'C' : np.random.randn(8),
                     'D' : np.random.randn(8)})
  
  print(df)
     Year    score         C         D
  0  2001  primary -1.277445  0.109920
  1  2002   second -1.814557  0.031803
  2  2001    third  0.246247  1.618513
  3  2002   fourth -1.042709  0.464520
  4  2001  primary  1.064651 -0.818139
  5  2002   second -0.197225  0.059878
  6  2001   fourth -0.292600 -0.432351
  7  2002    third  0.443334 -0.454662
  
  >>> df.groupby('Year')
  
  >>> df.groupby(['Year','score'])
  
  
  查看分组
  >>> df.groupby('Year').groups
  {'2001': Int64Index([0, 2, 4, 6], dtype='int64'), '2002': Int64Index([1, 3, 5, 7], dtype='int64')}
  
  分组打印
  for name,group in df.groupby(['Year','score']):
      print (name)
      print (group)
   
  ('2001', 'fourth')
     Year   score         C         D
  6  2001  fourth  0.240019  0.329059
  ('2001', 'primary')
     Year    score         C         D
  0  2001  primary  0.663303 -0.923613
  4  2001  primary -1.034212 -0.696587
  ('2001', 'third')
     Year  score         C         D
  2  2001  third  0.795344 -0.022841
  ('2002', 'fourth')
     Year   score         C         D
  3  2002  fourth -0.710015  2.105533
  ('2002', 'second')
     Year   score         C         D
  1  2002  second  0.701961  1.064701
  5  2002  second  0.796509 -0.582946
  ('2002', 'third')
     Year  score         C         D
  7  2002  third  0.849467 -0.248125
  
  
  

• 选择一个分组

  import pandas as pd
  import numpy as np
  
  df = pd.DataFrame({'Year' : ['2001', '2002', '2001', '2002',
                            '2001', '2002', '2001', '2002'],
                     'score' : ['primary', 'second', 'third', 'fourth',
                            'primary', 'second', 'fourth', 'third'],
                     'C' : np.random.randn(8),
                     'D' : np.random.randn(8)})
  
  
  grouped = df.groupby('Year')
  print (grouped.get_group('2001'))
  
     Year    score         C         D
  0  2001  primary  1.155751 -1.306120
  2  2001    third  0.029471 -0.231977
  4  2001  primary  0.553366  0.313114
  6  2001   fourth  0.698063 -1.472859
  

• 聚合

  聚合函数为每个组返回单个聚合值。当创建了分组(group by)对象，就可以对分组数据执行多个聚合操作。

  一个比较常用的是通过聚合或等效的agg方法聚合 -

  import pandas as pd
  import numpy as np
  
  df = pd.DataFrame({'Year' : ['2001', '2002', '2001', '2002',
                            '2001', '2002', '2001', '2002'],
                     'score' : ['primary', 'second', 'third', 'fourth',
                            'primary', 'second', 'fourth', 'third'],
                     'C' : np.random.randn(8),
                     'D' : np.random.randn(8)})
  
  
  grouped = df.groupby('Year')
  print (grouped['C'].agg(np.mean))
  

  • 聚合size方法

    import pandas as pd
    import numpy as np
    
    df = pd.DataFrame({'Year' : ['2001', '2002', '2001', '2002',
                              '2001', '2002', '2001', '2002'],
                       'score' : ['primary', 'second', 'third', 'fourth',
                              'primary', 'second', 'fourth', 'third'],
                       'C' : np.random.randn(8),
                       'D' : np.random.randn(8)})
    
    
    grouped = df.groupby('Year')
    print (grouped['C'].agg(np.size))
    

  • 同时聚合多种方法

    import pandas as pd
    import numpy as np
    
    df = pd.DataFrame({'Year' : ['2001', '2002', '2001', '2002',
                              '2001', '2002', '2001', '2002'],
                       'score' : ['primary', 'second', 'third', 'fourth',
                              'primary', 'second', 'fourth', 'third'],
                       'C' : np.random.randn(8),
                       'D' : np.random.randn(8)})
    
    
    grouped = df.groupby('Year')
    print (grouped['C'].agg([np.size,np.sum,np.mean]))
    
          size       sum      mean
    Year                          
    2001   4.0 -3.839661 -0.959915
    2002   4.0 -3.633480 -0.908370
    
    

• 转换

  分组或列上的转换返回索引大小与被分组的索引相同的对象。因此，转换应该返回与组块大小相同的结果。

  score = lambda x: (x - x.mean())
  

  实例

  import pandas as pd
  import numpy as np
  
  df = pd.DataFrame({'Year' : ['2001', '2002', '2001', '2002',
                            '2001', '2002', '2001', '2002'],
                     'score' : ['primary', 'second', 'third', 'fourth',
                            'primary', 'second', 'fourth', 'third'],
                     'C' : np.random.randn(8),
                     'D' : np.random.randn(8)})
  
  
  grouped = df.groupby('Year')
  score = lambda x: (x - x.mean())
  print(df)
  print(grouped['C'].agg(np.mean))
  print (grouped['C'].transform(score))
  
     Year    score         C         D
  0  2001  primary  1.455783  0.272765
  1  2002   second -0.891035  1.399249
  2  2001    third -0.335082  0.777773
  3  2002   fourth -0.938394 -0.614140
  4  2001  primary -0.770775  0.677035
  5  2002   second  1.561887  0.343912
  6  2001   fourth -0.426879 -0.551144
  7  2002    third  1.070326 -0.438464
  Year
  2001   -0.019239
  2002    0.200696
  Name: C, dtype: float64
  0    1.475021
  1   -1.091731
  2   -0.315844
  3   -1.139090
  4   -0.751537
  5    1.361191
  6   -0.407640
  7    0.869630
  Name: C, dtype: float64
  
          
          可以看到1.475021 = 1.455783 - （-0.019239）