Evolutionary Algorithm for overlapping community detection using a merged maximal cliques representation scheme

Abstract

Overlapping community detection in complex networks such as social, biological, economic, and other real-world networks has become an important area of research in the past two decades. Community detection problem can be modeled as a multiobjective optimization problem and it has been successfully solved using Evolutionary Algorithms (EA). The representation scheme of a solution in an EA affects the size of search space and thereby its convergence. A clique-based representation scheme is suitable to represent overlapping communities since overlapping cliques represent overlapping communities. The community to which a clique belongs becomes the community of its containing nodes. However, there are two issues viz., (1) cliques may heavily overlap since they might share many nodes with other cliques (2) one-node and two-node cliques increase the length of the representation of the EA. This paper proposes a merged-maximal-clique based representation scheme which reduces the chromosome length with fewer number of cliques. The proposed scheme overcomes the drawbacks of the existing clique-based scheme and naturally brings the initial solutions of the EA closer to the actual solutions. Moreover, operators of the EA are enhanced by renumbering the solutions to avoid duplicate solutions taking part in evolution. An EA based on decomposition equipped with the new representation scheme and enhanced operators is tested on both real-world and synthetic networks. The proposed EA finds better community partitions in fewer generations compared to the existing clique-based and non-clique-based algorithms. This study proves that clique-based schemes are efficient for overlapping community detection.