Optimization of K-means clustering method using hybrid capuchin search algorithm

This work presents Hybrid Capuchin Search Algorithm (HCSA) as a meta-heuristic method to deal with the vexing problems of local optima traps and initialization sensitivity of the K-means clustering technique. This study was inspired by the popularity and permanence of meta-heuristics in presenting convincing solutions, which sparked various efficient methods and computational tools to tackle difficult and practical real-world problems. The movement behavior of CSA is strengthened using the Chameleon Swarm algorithm to support the search agents of CSA to more effectively explore and exploit each potential region of the search space. This increases the capacity of both exploitation and exploration of the traditional CSA. Besides, the search agents of CSA utilize the rotation mechanism in CS to migrate to new spots outside the nearby regions to perform global search. This mechanism improves the search proficiency of CSA as well as the intensification and diversity abilities of the search agents. These expansion aptitudes of CSA expand its exploitation potential and broaden the range of search scopes, sizes, and directions in conducting clustering activities. A total of 16 different datasets from diverse sources, each with a different level of complexity, characteristics, and dimension, are used to assess the performance of the developed HCSA method on clustering tasks. According to the experimental results, the proposed HCSA performs statistically significantly better than the K-means clustering algorithm and eight meta-heuristics-based clustering in terms of both distance and performance metric measures.