在分类变量上旋转熊猫数据帧



我有一个包含分类变量的数据帧:

{'SysID': {0: '00721778',
1: '00721778',
2: '00721778',
3: '00721779',
4: '00721779'},
'SoftwareComponent': {0: 'AA13912',
1: 'AA24120',
2: 'AA21612',
3: 'AA30861',
4: 'AA20635'},
'SoftwareSubcomponent': {0: None,
1: 'AK21431',
2: None,
3: 'AK22116',
4: None}}

我想通过忽略任何 NULL 值来透视分类变量。零应该是填充物。输出应如下所示:

{'SysID': {0: '00721778', 1: '00721779'},
'SoftwareCom-AA13912': {0: '1', 1: '0'},
'SoftwareCom-AA24120': {0: '1', 1: '0'},
'SoftwareCom-AA21612': {0: '1', 1: '0'},
'SoftwareCom-AA30861': {0: '0', 1: '1'},
'SoftwareCom-AA20635': {0: '0', 1: '1'},
'SoftwareSub-AK21431': {0: '1', 1: '0'},
'SoftwareSub-AK22116': {0: '0', 1: '1'}}

怎么做?

您可以使用pd.crosstab(),然后在使用pd.concat()之前重命名数据帧列:

df1 = pd.crosstab(df['SysID'], df['SoftwareComponent'])
df1.columns = [df1.columns.name + '-' + i for i in df1.columns]
df2 = pd.crosstab(df['SysID'], df['SoftwareSubcomponent'])
df2.columns = [df2.columns.name + '-' + i for i in df2.columns]
final = pd.concat([df1, df2], axis=1)

收益 率:

SoftwareComponent-AA13912  SoftwareComponent-AA20635  
SysID                                                            
00721778                          1                          0   
00721779                          0                          1   
SoftwareComponent-AA21612  SoftwareComponent-AA24120  
SysID                                                            
00721778                          1                          1   
00721779                          0                          0   
SoftwareComponent-AA30861  SoftwareSubcomponent-AK21431  
SysID                                                               
00721778                          0                             1   
00721779                          1                             0   
SoftwareSubcomponent-AK22116  
SysID                                   
00721778                             0  
00721779                             1 

使用to_dict(),您可以返回:

{'SoftwareComponent-AA13912': {'00721778': 1, '00721779': 0}, 'SoftwareComponent-AA20635': {'00721778': 0, '00721779': 1}, 'SoftwareComponent-AA21612': {'00721778': 1, '00721779': 0}, 'SoftwareComponent-AA24120': {'00721778': 1, '00721779': 0}, 'SoftwareComponent-AA30861': {'00721778': 0, '00721779': 1}, 'SoftwareSubcomponent-AK21431': {'00721778': 1, '00721779': 0}, 'SoftwareSubcomponent-AK22116': {'00721778': 0, '00721779': 1}}

您可以在进行一些清理后使用pd.crosstab。我们将堆叠(这将忽略所有None值(并创建列名,因为您希望将SofwareCom和SoftwareSub视为相同。

import pandas as pd
df = df.set_index('SysID').stack().reset_index(level=1)
df['val'] = df['level_1'].str[0:11] + '-' + df[0]
pd.crosstab(df.index, df.val).rename_axis('SysID', 0).rename_axis(None,1).reset_index()

输出:

SysID  SoftwareCom-AA13912  SoftwareCom-AA20635  SoftwareCom-AA21612  SoftwareCom-AA24120  SoftwareCom-AA30861  SoftwareSub-AK21431  SoftwareSub-AK22116
0  00721778                    1                    0                    1                    1                    0                    1                    0
1  00721779                    0                    1                    0                    0                    1                    0                    1

如果你有可能有多个计数,并且只想要 1 和 0,那么你可以类型转换为布尔值,然后回到 int,或者只使用.clip

pd.crosstab(df.index, df.val).rename_axis('SysID', 0).rename_axis(None,1).clip(0,1).reset_index()

最新更新