Core Concepts
The author introduces a correlation-based fuzzy cluster validity index, the WP index, to accurately detect the optimal number of clusters and provide secondary options. The approach is based on the correlation between actual distances and adjusted centroids.
Abstract
The study introduces the WiroonsriโPreedasawakul (WP) index for cluster analysis, outperforming existing indexes in detecting optimal clusters. It remains effective even with high fuzziness parameters. The WP index offers multiple optimal choices for users, enhancing flexibility in selecting the final number of clusters. The study compares the WP index with various existing indexes across different datasets, showcasing its superior performance in accurately detecting clusters and providing secondary options.
Stats
The WP index outperforms most existing indexes in terms of accurately detecting the optimal number of clusters.
The study evaluates and compares the performance of the WP index with several existing indexes, including XieโBeni, PakhiraโBandyopadhyayโMaulik, Tang, WuโLi, generalized C, and Kwon2.