如何从嵌套字典创建扩展的熊猫数据框？

Question

我有一个嵌套的字典，并试图以此创建一个pandas数据框，但是它只提供了两列，我希望所有的字典键都是列。

MWE

import numpy as np
import pandas as pd

history = {'validation_0':
               {'error': [0.06725,0.067,0.067],
                'error@0.7': [0.104125,0.103875,0.103625],
                'auc': [0.92729,0.932045,0.934238],
               },
          'validation_1': 
              {'error': [0.1535,0.151,0.1505],
                'error@0.7': [0.239,0.239,0.239],
                'auc': [0.898305,0.905611,0.909242]
               }
          }


df = pd.DataFrame(history)
print(df)
                             validation_0                    validation_1
error             [0.06725, 0.067, 0.067]         [0.1535, 0.151, 0.1505]
error@0.7  [0.104125, 0.103875, 0.103625]           [0.239, 0.239, 0.239]
auc         [0.92729, 0.932045, 0.934238]  [0.898305, 0.905611, 0.909242]

必需

dataframe with following columns:
validation_0_error validation_1_error validation_0_error@0.7 validation_1_error@0.7  validation_0_auc validation_1_auc

Answer 1

您也可以在explode之后json_normalize：

print (pd.json_normalize(history).apply(pd.Series.explode).reset_index(drop=True))

  validation_0.error validation_0.error@0.7 validation_0.auc validation_1.error validation_1.error@0.7 validation_1.auc
0            0.06725               0.104125          0.92729             0.1535                  0.239         0.898305
1              0.067               0.103875         0.932045              0.151                  0.239         0.905611
2              0.067               0.103625         0.934238             0.1505                  0.239         0.909242

Answer 2

让我们尝试一下：

a = df.unstack()

pd.DataFrame(a.values.tolist(), index=a.index).T

如果您从history开始，

pd.concat({k:pd.DataFrame(v) for k,v in history.items()}, axis=1)

输出：

                      validation_0                     validation_1                    
         error  error@0.7      auc        error error@0.7       auc
0      0.06725  0.104125  0.927290       0.1535     0.239  0.898305
1      0.06700  0.103875  0.932045       0.1510     0.239  0.905611
2      0.06700  0.103625  0.934238       0.1505     0.239  0.909242