Cluster Validation by Measurement of Clustering Characteristics Relevant to the User
Christian Hennig · 2019
This chapter presents a range of cluster validation indexes that provide a multivariate assessment covering different complementary aspects of cluster validity. It focuses on “internal” validation criteria that measure the quality of a clustering without reference to external information such as a known “true” clustering. The chapter compares different clusterings on the same data, which is often referred to as “relative” cluster validation. This can be used to select one of a set of clusterings from different methods, or from the same method ran with different parameters such as different numbers of clusters. The chapter presents indexes measuring different relevant aspects of a clustering and defines an aggregated index that can be adapted to practical needs. It introduces a calibration scheme using randomly generated clusterings and applies the methodology to two data sets, one illustrative artificial one and a real data set regarding species delimitation.