A multi-objective optimization method for community detection using a novel heuristic search

Abstract Identifying structural patterns of complex networks is crucial to the study of community detection. The study of complex networks has recently been facilitated by several evolutionary methods. The main challenges of using evolutionary-based methods are their lower accuracies and their limitations in handling large networks. The evolutionary-based approaches treat most of these methods as single-objective tasks. Investigating a problem with multiple objectives can lead to a more precise identification of the community structure in a network. This is because each objective can capture distinct properties or aspects of the network, resulting in a more comprehensive understanding of its structure. In our paper, we introduce MOGGA+, which is a combination of a multi-objective genetic algorithm and local search strategies, aimed at tackling these issues. The local search strategy is primarily employed to accelerate convergence and enhance the precision of method. The solutions in MOCGA+ are represented using a vector-based method. The number of communities does not need to be known before the search begins because the utilization of this representation type diminishes the exploration range. Through experiments conducted on both LFR and real-world networks, the suggested method has demonstrated the ability to identify communities with higher accuracy while requiring fewer generations.