Multi-class multi-label confusion matrix with Sklearn

Question

I am working with a multi-class multi-label output from my classifier. The total number of classes is 14 and instances can have multiple classes associated. For example:

y_true = np.array([[0,0,1], [1,1,0],[0,1,0])
y_pred = np.array([[0,0,1], [1,0,1],[1,0,0])

The way I am making my confusion matrix right now:

matrix = confusion_matrix(y_true.argmax(axis=1), y_pred.argmax(axis=1))
print(matrix)

Which gives an output like:

[[ 79   0   0   0  66   0   0 151   1   8   0   0   0   0]
 [  4   0   0   0  11   0   0  27   0   0   0   0   0   0]
 [ 14   0   0   0  21   0   0  47   0   1   0   0   0   0]
 [  1   0   0   0   4   0   0  25   0   0   0   0   0   0]
 [ 18   0   0   0  50   0   0  63   0   3   0   0   0   0]
 [  4   0   0   0   3   0   0  19   0   0   0   0   0   0]
 [  2   0   0   0   3   0   0  11   0   2   0   0   0   0]
 [ 22   0   0   0  20   0   0 138   1   5   0   0   0   0]
 [ 12   0   0   0   9   0   0  38   0   1   0   0   0   0]
 [ 10   0   0   0   3   0   0  40   0   4   0   0   0   0]
 [  3   0   0   0   3   0   0  14   0   3   0   0   0   0]
 [  0   0   0   0   2   0   0   3   0   0   0   0   0   0]
 [  2   0   0   0  11   0   0  32   0   0   0   0   0   0]
 [  0   0   0   0   0   0   0   3   0   0   0   0   0   7]]

Now, I am not sure if the confusion matrix from sklearn is capable of handling multi-label multi-class data. Could someone help me with this?

Answer 1

What you need to do is to generate multiple binary confusion matrices (since essentially what you have are multiple binary labels)

Something along the lines of:

import numpy as np
from sklearn.metrics import confusion_matrix

y_true = np.array([[0,0,1], [1,1,0],[0,1,0]])
y_pred = np.array([[0,0,1], [1,0,1],[1,0,0]])

labels = ["A", "B", "C"]

conf_mat_dict={}

for label_col in range(len(labels)):
    y_true_label = y_true[:, label_col]
    y_pred_label = y_pred[:, label_col]
    conf_mat_dict[labels[label_col]] = confusion_matrix(y_pred=y_pred_label, y_true=y_true_label)


for label, matrix in conf_mat_dict.items():
    print("Confusion matrix for label {}:".format(label))
    print(matrix)

Answer 2

Now you can use (version 0.21) sklearn.metrics.multilabel_confusion_matrix

https://scikit-learn.org/stable/modules/generated/sklearn.metrics.multilabel_confusion_matrix.html

We try to predict two labels for each example

import sklearn.metrics as skm
y_true = np.array([
                [0,0], [0,1], [1,1], [0,1], [0,1], [1,1]
              ])
 y_pred = np.array([
                [1,1], [0,1], [0,1], [1,0], [0,1], [1,1] 
              ])

 cm = skm.multilabel_confusion_matrix(y_true, y_pred)
 print(cm)
 print( skm.classification_report(y_true,y_pred))

Confusion matrix for labels:

[[[2 2]
  [1 1]]

 [[0 1]
  [1 4]]]

Classification report:

              precision    recall  f1-score   support

         0       0.33      0.50      0.40         2
         1       0.80      0.80      0.80         5

micro avg        0.62      0.71      0.67         7
macro avg        0.57      0.65      0.60         7
weighted avg     0.67      0.71      0.69         7
samples avg      0.67      0.58      0.61         7