A High-Performance Evolutionary Multiobjective Community Detection Algorithm

Abstract: Community structure is a key feature of complex networks, underpinning a diverse range of phenomena across social, biological, and technological systems. While traditional methods, such as Louvain and Leiden, offer efficient solutions, they rely on single-objective optimization, often failing to capture the multifaceted nature of real-world networks. Multi-objective approaches address this limitation by considering multiple structural criteria simultaneously, but their high computational cost restricts their use in large-scale settings. We propose HP-MOCD, a high-performance, fully parallel evolutionary algorithm based on NSGA-II, designed to uncover high-quality community structures by jointly optimizing conflicting objectives. HP-MOCD leverages topology-aware genetic operators and parallelism to efficiently explore the solution space and generate a diverse Pareto front of community partitions. Experimental results on large synthetic benchmarks demonstrate that HP-MOCD consistently outperforms existing multi-objective methods in runtime, while achieving superior or comparable detection accuracy. These findings position HP-MOCD as a scalable and practical solution for community detection in large, complex networks.