Motif-based community detection in multiplex networks

Abstract

Community detection in multiplex networks usually relies on edge-based strategies. These strategies try to find an underlying community structure able to represent all the communities at different layers emerging from edges composing the network. Recent studies on community detection in complex networks have highlighted that network motifs, i.e. small groups of nodes interconnected in patterns occurring more frequently than in a random network, are able to reveal the organization of the network in motif-based communities. This paper proposes a motif-based community detection method that exploits many-objective optimization. The focus of the method is to find a clustering on a multiplex network that maximizes the number of instances of a motif inside the same community, while minimizing cutting instances of the same motif on all layers. The method employs a many-objective genetic algorithm that simultaneously optimizes the concept of motif conductance on all the layers. Simulations on several real-world traces show the superiority of our method with respect to existing motif-based methods for single-layer networks.