熔化函数Pandas的InvalidIndexError

Question

我有以下数据帧，当我应用熔融函数时：

numerics = ['int16', 'int32', 'int64', 'float16', 'float32', 'float64']
df_rm_features_melted = df_rm_features.melt(
id_vars=['id', 'date'],
value_vars=df_rm_features.select_dtypes(include=numerics).columns
)

我得到错误：InvalidIndexError: Reindexing only valid with uniquely valued Index objects

id      date  factoring  lglc overdue_60  distress  max_dpd  tr_op
0288     12/1/2018  0       1       1           1        0       0
0288     1/1/2019   0       0       0           0        10      1
0288     2/1/2019   0       0       0           0        2       1
0288     3/1/2019   0       0       0           0        52      1
0288     4/1/2019   0       0       0           1        2       0

Answer 1

首先，我从您的代码中得到的错误是ValueError: arrays must all be same length。可能是因为id在id_vars列表中，也是一个数字列，所以最终也在value_vars列表中。

要从value_vars中删除id和任何其他列，而不必显式指定，请将numpy.setdiff1d()与select_dtypes子句一起使用：

id_vars=['id', 'date']
wanted_vals = df_rm_features.select_dtypes(include=numerics).columns
canhave_vals = np.setdiff1d(wanted_vals, id_vars)
df_rm_features.melt(id_vars=id_vars,
value_vars=canhave_vals)

输出：

id       date    variable  value
0   288  12/1/2018    distress      1
1   288   1/1/2019    distress      0
2   288   2/1/2019    distress      0
3   288   3/1/2019    distress      0
4   288   4/1/2019    distress      1
5   288  12/1/2018   factoring      0
6   288   1/1/2019   factoring      0
7   288   2/1/2019   factoring      0
8   288   3/1/2019   factoring      0
9   288   4/1/2019   factoring      0
10  288  12/1/2018        lglc      1
11  288   1/1/2019        lglc      0
12  288   2/1/2019        lglc      0
13  288   3/1/2019        lglc      0
14  288   4/1/2019        lglc      0
15  288  12/1/2018     max_dpd      0
16  288   1/1/2019     max_dpd     10
17  288   2/1/2019     max_dpd      2
18  288   3/1/2019     max_dpd     52
19  288   4/1/2019     max_dpd      2
20  288  12/1/2018  overdue_60      1
21  288   1/1/2019  overdue_60      0
22  288   2/1/2019  overdue_60      0
23  288   3/1/2019  overdue_60      0
24  288   4/1/2019  overdue_60      0
25  288  12/1/2018       tr_op      0
26  288   1/1/2019       tr_op      1
27  288   2/1/2019       tr_op      1
28  288   3/1/2019       tr_op      1
29  288   4/1/2019       tr_op      0