根據條件在groupby數據框上創建匯總表

Question

我有一個看起來像的熊貓數據框df

userid  trip_id segmentid   actual  prediction
  1       13       40          3       3
  1       6        2           1       1
  1       44       3           2       3
  2       70       19          1       1
  2       12       5           0       0

我需要創建一個匯總數據框dfsummary ，該數據框按列userid分組 ，具有三列userid，correct_classified，corrected_classified。 如果實際值和預測值相同，則將其正確分類，否則將分類錯誤。

我可以將對整個數據框的correct_classfied視為

correct_classified = submission[(submission['Actual'] == submission['prediction'])]
incorrect_classified = submission[(submission['Actual'] != submission['prediction'])]

但是不知道創建按用戶ID分組的摘要表的想法，它應該像這樣

userid  correct_classified  incorrect_classified
  1             2                    1
  2             2                    0

Answer 1

您可以在創建條件數組后使用pd.crosstab ：

flags = np.where(df['actual'].eq(df['prediction']), 'correct', 'incorrect')

res = pd.crosstab(df['userid'], flags)

print(res)

col_0   correct  incorrect
userid                    
1             2          1
2             2          0

Answer 2

您也可以使用pivot table

m = df['actual']==df['prediction']

# assign the conditions to new columns and aggregate.  
df.assign(correct_classified=m,incorrect_classified=~m).pivot_table(index='userid',
                                                                    aggfunc='sum',
                                                                    values=['correct_classified',
                                                                            'incorrect_classified'])

輸出：

     correct_classified  incorrect_classified
userid                                          
1                      2.0                   1.0
2                      2.0                   0.0