Source: google-ml-course

Confusion matrix

• $N \times N$ matrix which shows how successful the classification model’s predictions were.
• $N$: number of classes, e.g. $N=2$ for true/false classes
• Axes: predicted class/label, actual class/label

In multi-class classification, the confusion matrix can help to identify mistake patterns—does the model tend to mistakenly predict a certain class for another?