Skip to content

Independent standards discovery and compliance platform

How we work

ISO/IEC TS 4213:2022 - Machine Learning Classifier Evaluation

This Technical Specification focuses on measuring the performance of machine learning classification. It is relevant when a team needs to justify its choice of performance measures or compare classifiers whose results are reported on different datasets or across different output classes.

What you are buying

Before comparing scores, record the classes, the data used for evaluation and which types of mistakes have practical consequences. That helps reviewers understand what a reported measure captures and why a higher overall score may leave an important performance question unresolved.

Frequently asked questions

Can I use this specification when comparing classifiers from two suppliers?

Yes. Compare results under a common evaluation setup. Check that the suppliers use comparable datasets, labels and performance measures before drawing a conclusion.

Is this the same document as a method for evaluating regression?

Its scope is machine learning classification performance. For a model that predicts a continuous value, choose evaluation measures appropriate to regression and the intended use.