python - 如何将一系列索引/类别转换为分类数组

Question

如何将一系列索引转换为二维数组，该数组表示由列表中的索引值定义的类别/分类器？

例如:

import numpy as np

aList = [0,1,0,2]
anArray = np.array(aList)

resultArray = convertToCategories(anArray)

convertToCategories() 的返回值如下:

[[1,0,0],             # the 0th element of aList is index category 0
[0,1,0],              # the 1st element of aList is index category 1
[1,0,0],              # the 2nd element of aList is index category 0
[0,0,1]]              # the 3rd element of aList is index category 2

最后，我当然可以:

• 解析列表，
• 计算类别的数量(它是连续的/连续的，可以简单地找到最大值)
• 创建一个大小合适的归零数组
• 然后重新解析列表，根据列表给定的索引用1(或True)填充数组。

但我想知道是否存在更 pythonic 或专用的 numpy 或 pandas 函数来实现这种转换。

Answer 1

你可以这样做 -

import numpy as np

# Size parameters
N = anArray.size
M = anArray.max()+1

# Setup output array
resultArray = np.zeros((N,M),int)

# Find out the linear indices where 1s would be put
idx =  (np.arange(N)*M) + anArray

# Finally, put 1s at those places for the final output
resultArray.ravel()[idx] = 1

sample 运行-

In [188]: anArray
Out[188]: array([0, 1, 0, 2, 4, 1, 3])

In [189]: resultArray
Out[189]: 
array([[1, 0, 0, 0, 0],
       [0, 1, 0, 0, 0],
       [1, 0, 0, 0, 0],
       [0, 0, 1, 0, 0],
       [0, 0, 0, 0, 1],
       [0, 1, 0, 0, 0],
       [0, 0, 0, 1, 0]])

或者，最好直接使用行和列索引索引到输出数组中 -

# Setup output array and put 1s at places indexed by row and column indices.
# Here, anArray would be the column indices and [0,1,....N-1] would be the row indices 
resultArray = np.zeros((N,M),int)
resultArray[np.arange(N),anArray] = 1