How to evaluate clustering with ground truth?
Artificial Intelligence
2026-06-25 v1
Abstract
External indexes can be used for cluster evaluation when ground truth is available. We review the most common external validity indexes focusing on set-matching-based measures. We recommend centroid index (CI), because it is an intuitive cluster-level measure with an explainable result. If we need a more fine-tuned, point-level measure, there are more choices. Pair-set index (PSI) provides a normalized score which is not biased by cluster sizes. If all points should matter equally, then clustering accuracy (ACC) or any other set-matching measure is suitable.
Cite
@article{arxiv.2606.27061,
title = {How to evaluate clustering with ground truth?},
author = {Pasi Fränti},
journal= {arXiv preprint arXiv:2606.27061},
year = {2026}
}
Comments
Preprint of a book chapter to appear: P. Fränti, "How to evaluate clustering with ground truth?", In Center-based clustering, Springer Nature, 2026