将元组列表转换为 3d numpy 数组

Question

我有一个数据集，它是一个元组列表，我需要将其转换为 3d numpy 数组。举个例子：

       data = [(1, 65, -18, -1, -1 ),
       (1, -18,-1, -1,-1),
      (2, 65, -19, -1, -1),
      (2, 65, -18, -1, -1),
       (3, 62, -18, -1, -1)]

我想像这样创建一个 3d numpy 数组：

     array[[[[65], [-18], [-1], [-1]],
          [[-18], [-1], [-1], [-1]]],
          [[[65], [-19], [-1], [-1]],
          [[65], [-18], [-1], [-1]]],
          [[[62], [-18], [-1], [-1]]]]

  

   

如何使用 Numpy 库实现此目的？

Answer 1

遍历每个值，并重构你的列表：

data = [(1, 65, -18, -1, -1 ),
   (1, -18,-1, -1),
   (2, 65, -19, -1, -1),
   (2, 65, -18, -1, -1),
   (3, 62, -18, -1, -1)]

[[[[val] for val in row[1:]] for row in data]]

输出：

[[[[65], [-18], [-1], [-1]],
[[-18], [-1], [-1]],
[[65], [-19], [-1], [-1]],
[[65], [-18], [-1], [-1]],
[[62], [-18], [-1], [-1]]]]

Answer 2

我认为这样做并存储在 numpy 数组中的一种简短方法是

import numpy as np

data = [(1, 65, -18, -1, -1 ),
        (1, 65, -18,-1, -1),
        (2, 65, -19, -1, -1),
        (2, 65, -18, -1, -1),
        (3, 62, -18, -1, -1)]

numpy_data = np.array(data)[:,1:].reshape((5,4,1))
numpy_data

输出

array([[[ 65],
        [-18],
        [ -1],
        [ -1]],

       [[ 65],
        [-18],
        [ -1],
        [ -1]],

       [[ 65],
        [-19],
        [ -1],
        [ -1]],

       [[ 65],
        [-18],
        [ -1],
        [ -1]],

       [[ 62],
        [-18],
        [ -1],
        [ -1]]])

你可以随心所欲地操作

Answer 3

应用

np.unique

+

np.split

的组合，根据初始

arr

的第一列中包含的“id”获得数组组：

arr = np.asarray(data)
groups = np.split(arr[:, 1:], np.unique(arr[:, 0], return_index=True)[1])[1:]
groups = [a.reshape(-1, a.shape[1], 1) for a in groups]
print(groups)

---

[array([[[ 65],
        [-18],
        [ -1],
        [ -1]],

       [[-18],
        [ -1],
        [ -1],
        [ -1]]]), array([[[ 65],
        [-19],
        [ -1],
        [ -1]],

       [[ 65],
        [-18],
        [ -1],
        [ -1]]]), array([[[ 62],
        [-18],
        [ -1],
        [ -1]]])]