在sklearn指标confusion_matrix中包括零命中的行和列

Question

I want to combine results from several confusion matrixes in a single matrix. My prediction and ground truth values can be 0, 1, 2. I can add up values whenever all values are present. However the number of rows and columns changes when some values are not present at all. Hence, I cannot add such array to the resulting array. How to force confusion matrix include rows and columns with zero samples in them?

import numpy as np
from sklearn.metrics import confusion_matrix

combined_confusion_matrix = np.zeros((3,3))
gt1 = np.array([0,1,2,0,1,2])
pred1 = np.array([2,1,2,0,1,2])
cm1 = confusion_matrix(gt1, pred1)  # [[1 0 1][0 2 0][0 0 2]]
combined_confusion_matrix +=  cm1
gt2 = np.array([0,0,2,2])
pred2 = np.array([0,2,2,2])
cm2 = confusion_matrix(gt2, pred2)  # Got [[1 1][0 2]], Desired [[1 0 1][0 0 0][0 0 2]]
combined_confusion_matrix += cm2  # Error due to different dimensions

Answer 1

This worked for me:

my_bins = [0, 1, 2]
cm2 = confusion_matrix(gt2, pred2, labels=my_bins)