useR! 2024

Selecting the appropriate number of clusters is a critical step in applying clustering methods. To assist in this process, various cluster validity indices (CVIs) have been developed. These indices are designed to identify the optimal number of clusters within a dataset. However, users may not always seek the absolute optimal number of clusters but rather a secondary option that better aligns with their contexts. This realization has led us to introduce a Bayesian cluster validity index (BCVI), which builds upon existing indices. The BCVI utilizes a Dirichlet prior, resulting in the same posterior distribution. We evaluate BCVI using the Wiroonsri index for hard clustering and the WP index for soft clustering as underlying indices. We compare the performance of BCVI with that of the original underlying indices and several other existing CVIs, including DB, STR, XB, and KWON2 indices. Our BCVI offers clear advantages in situations where users can specify their desired range for the final number of clusters. Additionally, we showcase the practical applicability of our approach through MRI images. These tools are also published as a new R package `BayesCVI' available on CRAN.