Multi-Objective Evolutionary Algorithm with Decomposition for Enhanced Community Detection in Signed Networks

Abstract

In complex networks, community detection is an NP-hard problem that usually requires solving two subproblems: the problem of generating algorithms for network partitioning and the problem of evaluating the quality of partitioning. In this study, we propose a novel community detection model based on a decomposition-based multi-objective evolutionary algorithm (MOEA/D) called ‘Multi-Objective Community Detection Model based on Scoring (MOCDR)’.MOCDR improves the community detection performance by optimising the scoring of both intra-community (internal connectivity) and inter-community (external connectivity) types of connectivity. detection performance. To enhance the search capability of MOEA/D, we propose a new mutation operator, E-Mutate, which strikes a better balance between local and global search. Experimental results show that MOCDR performs well in scoring internal and external connections, improving the accuracy and efficiency of community detection.