Enhancing Community Detection: A Comparative Study of Metaheuristic Optimization Techniques

Abstract

Community detection in networks is a crucial task across diverse fields. While modularity maximization is a widely used approach, it suffers from limitations such as resolution limits and sensitivity to noise. Modularity density, an alternative measure, addresses some of these issues by focusing on minimizing out-of-cluster links. Building upon previous work on Modified Modularity Density Maximization (MMDM), which minimizes the deepest out-of-cluster connection, this paper investigates the application of advanced metaheuristic optimization techniques to enhance community detection performance. We compare the performance of Genetic Algorithms (GA), Particle Swarm Optimization (PSO), and Tabu Search in solving the MMDM problem. These algorithms are evaluated against the previously proposed Density Ratio (DR) Heuristic, a method designed to handle larger datasets where exact solutions are computationally prohibitive. Our results demonstrate the effectiveness of these metaheuristics in finding high-quality solutions to the MMDM problem, and we provide a comparative analysis of their performance in terms of solution quality, computational efficiency, and scalability across various network datasets. This study contributes to the advancement of community detection techniques by exploring the potential of metaheuristics for solving complex optimization problems in network analysis.