Selecting External Validation Measures for K-means Clustering

Abstract

Cluster validity is a long standing challenge in the clustering literature. While many evaluation measures have been developed for cluster validity, these measures often provide inconsistent information about the clustering performance, and the best suitable measures to use remain unclear in practice. Our study in this chapter fills this crucial void by giving an organized study of sixteen external validation measures for K-means clustering. Specifically, we first propose a filtering criterion based on the uniform effect of K-means, and apply it for the identification of defective measures.