MAEA-CD: A Novel Multiagent-Based Evolutionary Algorithm for Community Structure Detection in Social Networks

Due to the substantial growth in user numbers and the rapid generation and dissemination of information on social networks, the study and analysis of these networks have become increasingly important. One of the critical problems that social network analysts focus on is the identification of communities/subgroups that may exist within these networks. Identifying these groups has numerous applications in marketing, advertising, and search engine optimization. This paper proposes a method named MAEA-CD, which combines distributed evolutionary algorithms with multiagent systems for the purpose of identifying communities in social networks. The main advantages of MAEA-CD lie in the simultaneous utilization of the capabilities of evolutionary algorithms—exploration and exploitation—and the features of multiagent systems—competition and cooperation, autonomy, and learning. This combination enables MAEA-CD to significantly alleviate the issues of premature convergence and entrapment in local optima that are prevalent in most conventional evolutionary algorithms. Experimental results across diverse community detection instances show that MAEA-CD is efficient in identifying social network communities. Its performance, on average, is 16.63% better on real-world benchmark networks, 25.30% better on GN synthetic benchmark networks, and 30.89% better on LFR synthetic benchmark networks compared to the mean performance of all baseline algorithms, averaged across all networks. To demonstrate broader engineering relevance beyond social networks, we present a case study on the U.S. Western States Power Grid. MAEA-CD uncovers operationally coherent partitions aligned with design and resilience objectives (e.g., controlled islanding and regional operation), illustrating its applicability to complex engineered infrastructure.