confusion_matrix() | ValueError: Classification metrics can't handle a mix of multiclass and multiclass-multioutput targets

Question

肯定有人问过，但我没有成功分析其他帖子的解决方案以解决我自己的这个问题。

我有很多分类模型，我想使用 confusion_matrix()

进行比较

matrix = confusion_matrix(y_test, y_pred) # ERROR

>>> y_pred
[[2 2 2 ... 2 2 2]
 [2 2 2 ... 2 2 2]
 [2 2 2 ... 2 2 2]
 ...
 [3 3 2 ... 3 2 3]
 [2 2 2 ... 2 2 2]
 [3 3 3 ... 3 3 3]]

>>> y_pred.shape
(500, 256)

>>> y_test
[1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1
 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1
 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1
 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1
 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2
 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2
 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2
 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2
 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 3 3 3 3
 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3
 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3
 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3
 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3
 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3]

>>> y_test.shape
(500, )

错误：

ValueError: Classification metrics can't handle a mix of multiclass and multiclass-multioutput targets

当对 y_pred 执行 .flatten() - 即一维数组 (500 * 256 = 128000):

ValueError: Found input variables with inconsistent numbers of samples: [500, 128000]

Answer 1

混淆矩阵的工作原理是比较每个预测值和实际值。无法将 1 与 [2,2,2....2,2,2]

进行比较

在你的例子中，你的 y_pred 是 2d 而你的 y_test 是 1d，这就是实际错误的来源。我相信您必须选择预测列表中最常见的数字。喜欢来自 [2,2,2....2,2]

的 2

所以这里是解决方案：

from scipy import stats 
import numpy as np

#taking the most frequent element from the predicted list
y_pred_list = [int(stats.mode(arr)[0]) for arr in y_pred.tolist()] #convert to list

y_pred_array = np.array(y_pred_list)  #convert to 1D with same shape of y_test

print(y_pred_array.shape)

print(y_pred_array)

matrix = confusion_matrix(y_test, y_pred_array)

confusion_matrix() | ValueError: Classification metrics can't handle a mix of multiclass and multiclass-multioutput targets

confusion_matrix() | ValueError: Classification metrics can't handle a mix of multiclass and multiclass-multioutput targets

python

classification

neural-network

confusion-matrix

multilabel-classification