Improved fuzzy C-means clustering algorithm based on fuzzy particle swarm optimization for solving data clustering problems

Abstract

The fuzzy c-means (FCM) clustering algorithm is adversely affected by its sensitivity to initial values and its low clustering accuracy. To mitigate these shortcomings, we proposed an improved fuzzy particle swarm optimization-fuzzy C-Means (IFPSO-FCM) algorithm to resolve the data-clustering challenges. In this algorithm, key enhancements included initializing clustering centers using Mahalanobis distances to alleviate the sensitivity to initial values. An objective function based on both inter- and intra-cluster evaluations was proposed to address the premature convergence. A modified particle swarm algorithm was designed to optimize the clustering centers. The proposed algorithm was applied to analyze the IRIS and WINE datasets, as well as to cluster and segment classical test images. The results indicated that the algorithm improved the stability of the analysis results while preserving high clustering accuracy and convergence speed, achieving an excellent performance compared with existing methods. Moreover, it exhibited superior performance in the analysis of fuzzy multi-shadow gray images.