如何使用实际数据帧中两列中的值索引另一个数据帧，从而在实际数据框中创建列

Question

美好的一天。我有两个数据集（df1，df2）。我试图通过使用列'site_before'作为df1中的行索引和列'site'作为df1中的列索引来填充df2中的列'值'。

数据集df1：

      ANA01  PHO01  ATL  BAL12  BOS07   
ANA01   0     0      3     3     3
PHO01   0     0      3     3     3
ATL    -3    -3      0     0     0 
BAL12   -3   -3      0     0     0 
BOS07   -3    -3     0     0     0

“第一列是行的索引”

数据集df2：

    Game_ID     site_before  site   values
1   ANA199804010    ANA01   ANA01   
3   ANA199804020    ANA01   ATL 
5   ANA199804030    ANA01   BAL12   
7   ANA199804040    ANA01   BOS07   
9   ANA199804050    ANA01   ANA01   
674 BOS199804300    BOS07   BOS07   
31  ANA199805010    BOS07   ANA01   
33  ANA199805020    PHO01   ANA01   
35  ANA199805030    PHO01   PHO01   
37  ANA199805040    PHO01   ATL 
39  ANA199805050    PHO01   BAL12

我试着这样做：

df2['values'] = df1.loc[df2['site_before'], df2['site']].values

但我得到一个错误，因为ValueError：错误的项目数量通过4864，放置意味着1

我期待的结果是：

    Game_ID     site_before site    values
1   ANA199804010    ANA01   ANA01   0
3   ANA199804020    ANA01   ATL     3
5   ANA199804030    ANA01   BAL12   3
7   ANA199804040    ANA01   BOS07   3
9   ANA199804050    ANA01   ANA01   0
674 BOS199804300    BOS07   BOS07   0
31  ANA199805010    BOS07   ANA01   -3
33  ANA199805020    PHO01   ANA01   0
35  ANA199805030    PHO01   PHO01   0
37  ANA199805040    PHO01   ATL     3
39  ANA199805050    PHO01   BAL12   3

Answer 1

使用DataFrame.join与MultiIndex Series创建的新DataFrame.stack：

df2 = df2.join(df1.stack().rename('new').rename_axis(('site_before','site')), 
               on=['site_before','site'])
print (df2)
          Game_ID site_before   site  new
1    ANA199804010       ANA01  ANA01    0
3    ANA199804020       ANA01    ATL    3
5    ANA199804030       ANA01  BAL12    3
7    ANA199804040       ANA01  BOS07    3
9    ANA199804050       ANA01  ANA01    0
674  BOS199804300       BOS07  BOS07    0
31   ANA199805010       BOS07  ANA01   -3
33   ANA199805020       PHO01  ANA01    0
35   ANA199805030       PHO01  PHO01    0
37   ANA199805040       PHO01    ATL    3
39   ANA199805050       PHO01  BAL12    3

替代方案是使用DataFrame.melt与DataFrame.merge和左连接：

df3 = df1.rename_axis('site_before').reset_index().melt('site_before', var_name='site')

---

df2 = df2.merge(df3, how='left')
print (df2)
         Game_ID site_before   site  new
0   ANA199804010       ANA01  ANA01    0
1   ANA199804020       ANA01    ATL    3
2   ANA199804030       ANA01  BAL12    3
3   ANA199804040       ANA01  BOS07    3
4   ANA199804050       ANA01  ANA01    0
5   BOS199804300       BOS07  BOS07    0
6   ANA199805010       BOS07  ANA01   -3
7   ANA199805020       PHO01  ANA01    0
8   ANA199805030       PHO01  PHO01    0
9   ANA199805040       PHO01    ATL    3
10  ANA199805050       PHO01  BAL12    3