Performance Analysis of Metaheuristic Algorithms for Solving Data Clustering Problems

Metaheuristic algorithms are powerful techniques that are widely adopted by research community to compute the optimal solutions for complex and diverse optimization problems. Meta-heuristic algorithms provide more flexibility compared to traditional clustering algorithm and more suitable to handle high-dimensional, noisy, or irregular datasets. This work investigates efficacy of the several state of art metaheuristic algorithms like K-Means, Particle swarm optimization (PSO), Ant colony optimization (ACO), Cat swarm optimization (CSO), Teaching-learning-based optimization (TLBO), Bat algorithm (BA) and Chemical reaction algorithm (CRO) for handling clustering problems. These algorithms efficiency is evaluated using intra cluster distance and f-measure parameters. For experimental analysis, Iris, Cancer, CMC, Wine, and Glass datasets are considered and findings stated that a single meta-heuristic algorithm cannot outperform with all datasets. It is also found that exploration and exploitation trade-off have significant impact on the performance of the meta-heuristic algorithms. The tuning of the user-defined parameters also affects the performance of the meta-heuristic algorithm.