python - 将来自先前数据的信息分组到向量的数据帧中

在PANDAS中是否有一个函数可以为序列的每个点创建一个向量，该向量具有N的值-一组列的序列的先前值？

Ej：

从

pd.DataFrame([('A1', 'B1', 'C1', 'X'),
              ('A2', 'B2', 'C2', 'X'),
              ('A3', 'B3', 'C3', 'X'),
              ('A4', 'B4', 'C4', 'X'),
              ('A5', 'B5', 'C5', 'X'),
              ('A6', 'B6', 'C6', 'X'),
              ('A7', 'B7', 'C7', 'X'),
              ('A8', 'B8', 'C8', 'X'),
              ('A9', 'B9', 'C9', 'X')])

至

pd.DataFrame([(['A1', None, None], ['B1', None, None], ['C1', None, None], 'X'),
              (['A2', 'A1', None], ['B2', 'B1', None], ['C2', 'C1', None], 'X'),
              (['A3', 'A2', 'A1'], ['B3', 'B2', 'A1'], ['C3', 'C2', 'C1'], 'X'),
              (['A4', 'A3', 'A2'], ['B4', 'B3', 'B2'], ['C4', 'C3', 'C2'], 'X'),
              (['A5', 'A4', 'A3'], ['B5', 'B4', 'B3'], ['C5', 'C4', 'C3'], 'X')])

最佳答案

我认为使用Numpy的stride_tricks库最方便

from numpy.lib.stride_tricks import as_strided as strided

v = df.values[:, :-1]

pad = np.empty((2, n1), v.dtype)
pad.fill(None)

v = np.row_stack([pad, v])

s0, s1 = v.strides
n0, n1 = v.shape

idx = df.index
col = df.columns[:-1]

pd.DataFrame(
    strided(v, (n0 - 2, n1, 3), (s0, s1, s0))[:, :, ::-1].tolist(),
    idx, col
).join(df.drop(col, 1))

                  0                 1                 2  3
0  [A1, None, None]  [B1, None, None]  [C1, None, None]  X
1    [A2, A1, None]    [B2, B1, None]    [C2, C1, None]  X
2      [A3, A2, A1]      [B3, B2, B1]      [C3, C2, C1]  X
3      [A4, A3, A2]      [B4, B3, B2]      [C4, C3, C2]  X
4      [A5, A4, A3]      [B5, B4, B3]      [C5, C4, C3]  X
5      [A6, A5, A4]      [B6, B5, B4]      [C6, C5, C4]  X
6      [A7, A6, A5]      [B7, B6, B5]      [C7, C6, C5]  X
7      [A8, A7, A6]      [B8, B7, B6]      [C8, C7, C6]  X
8      [A9, A8, A7]      [B9, B8, B7]      [C9, C8, C7]  X

关于python - 将来自先前数据的信息分组到向量的数据帧中，我们在Stack Overflow上找到一个类似的问题：https://stackoverflow.com/questions/53268527/