python - 如何 reshape 多索引数据框

Question

我有一个多索引 DataFrame，我想将它的列用作行，然后将该列重命名为度量的名称。

arrays = [np.array(['bar', 'bar', 'baz', 'baz', 'foo', 'foo', 'qux', 'qux']),
          np.array(['one', 'two', 'one', 'two', 'one', 'two', 'one', 'two'])]
df = pd.DataFrame(np.random.randn(8, 4), index=arrays)

                0         1         2         3
bar one  0.049409  0.533468  0.528360 -1.437937
    two  2.081377 -0.945535  0.237531 -0.781147
baz one  0.005216  1.158222 -1.178232 -1.470667
    two -0.043834 -0.320864 -1.568357  0.803620
foo one -0.758539 -1.009726  0.139992  0.281034
    two -1.806000  0.206872 -0.728195  1.051045
qux one -1.106591 -0.621868 -1.139649 -0.185527
    two  0.176220 -0.961532  3.587891  0.627658

我想让我的数据框看起来像这样:

            measure_name
bar one  0  0.049409
    two  1  -0.945535
    one  2  0.528360
    two  3  -0.781147

我不知道该怎么做。我尝试了 pd.melt() 但它摆脱了多索引，我需要有一种方法将列中的值绑定(bind)到索引。

提前致谢!

Answer 1

IIUC:

df.stack().to_frame('measure_name')

           measure_name
bar one 0      0.562183
        1      2.090766
        2     -0.164342
        3      0.499693
    two 0     -0.174269
        1     -0.997726
        2      0.820774
        3      0.243022
baz one 0     -0.158621
        1      0.520945
        2     -0.356393
        3      0.465289
    two 0     -1.187833
        1      0.886986
        2      1.415511
        3      0.940117
foo one 0     -0.010860
        1      0.126255
        2      1.131045
        3     -0.899853
    two 0     -1.121544
        1     -0.327184
        2      0.074396
        3      0.214501
qux one 0     -0.028317
        1     -1.476114
        2      1.415711
        3     -0.355655
    two 0      0.285167
        1      1.535384
        2      0.074326
        3     -1.860993