[英]confusion_matrix() | ValueError: Classification metrics can't handle a mix of multiclass and multiclass-multioutput targets
Definitely has been asked before, but I've not been successful at analysing other posts' solutions for my own instance of this problem.以前肯定有人问过,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的实例。
I have many classification models I want to compare using confusion_matrix()
我有许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, )
Error:错误:
ValueError: Classification metrics can't handle a mix of multiclass and multiclass-multioutput targets
When .flatten()
is performed on y_pred
- ie 1D array (500 * 256 = 128000):当
.flatten()
在y_pred
上执行时 - 即一维数组(500 * 256 = 128000):
ValueError: Found input variables with inconsistent numbers of samples: [500, 128000]
Confusion matrix works on the basis of comparision between each predicted value and actual value.混淆矩阵基于每个预测值与实际值之间的比较来工作。 It is impossible compare
1
with [2,2,2....2,2,2]
不可能将
1
与[2,2,2....2,2,2]
进行比较
In your case, your y_pred is 2d but your y_test is 1d, thats where the actual error came.在您的情况下,您的 y_pred 是 2d 但您的 y_test 是 1d,这就是实际错误出现的地方。 I believe that you have to choose the most common number in your predicted list.
我相信你必须在你的预测列表中选择最常见的数字。 Like
2
from [2,2,2....2,2]
像
[2,2,2....2,2]
2
的 2
So here is the solution: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)
声明:本站的技术帖子网页,遵循CC BY-SA 4.0协议,如果您需要转载,请注明本站网址或者原文地址。任何问题请咨询:yoyou2525@163.com.