A High-Performance Evolutionary Multiobjective Community Detection Algorithm

2025-06-02

AI Summary

Traditional single-objective community detection methods (e.g., Louvain, Leiden) fail to capture network multifacetedness, while existing multi-objective approaches suffer from high computational overhead and poor scalability to large-scale networks. To address this, we propose HP-MOCD, an efficient parallel multi-objective community detection framework. Built upon NSGA-II, HP-MOCD introduces novel topology-aware crossover and mutation operators and employs a hybrid MPI+OpenMP parallel architecture. It simultaneously optimizes conflicting objectives—including modularity, separation, and cohesion—while significantly enhancing both diversity and convergence speed of the Pareto front. On large synthetic benchmarks, HP-MOCD achieves 3.2–5.8× speedup over state-of-the-art methods, attaining state-of-the-art or comparable community quality. Notably, it is the first multi-objective method to enable practical community detection on million-node networks.

Technology Category

• Constraint Satisfaction and Optimization: Distributed CSP/Optimization
• Search and Optimization: Mixed Discrete/Continuous Search
• Data Mining & Knowledge Management: Graph Mining, Social Network Analysis & Community

Application Category

• Graph Algorithms and Modeling for the Web: Representation, reconstruction, and subgraph or motif discovery in Web-related graphs
• Web Mining and Content Analysis: Community question answering
• Economics, Online Markets and Human Computation: Incentives in network design for Web infrastructures and ecosystems

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.

Problem

• Detects communities in complex networks using multi-objective optimization
• Addresses high computational cost of existing multi-objective methods
• Improves scalability and accuracy for large network analysis

Innovation

• Evolutionary algorithm based on NSGA-II
• Topology-aware genetic operators
• Fully parallel for large-scale networks