Confusion matrices are used to evaluate potential weaknesses in classification methods. For a method designed to distinguish N classes, the confusion matrix as a size of  N × N. The numeric value at an element (i, j) denotes how often the method classified an instance of class i as an instance of class j. An ideal method achieves a perfect diagnoal matrix. Off-diagonal entries instead (i.e. cases where for any i≠j the numeric value at this element is nonzero) indicate a risk of the method confusing the classes i and j, and quantify the occurrence.

See also:

• true positive
• true negative
• false positive
• false negative

ISO/IEC TR 29119-11:2020 table used to describe the performance of a classifier (3.1.21) on a set of test data (3.1.75) for which the true and false values are known

ISTQB - CTAI Syllabus A technique for summarizing the ML functional performance of a classification algorithm