A differential evolution-based approach for community detection in multilayer networks with attributes

A differential evolution based algorithm for detecting community structure in multilayer networks with node attributes is proposed. The method optimizes a fitness function that combines structural connectivity of each layer with node similarity to obtain multilayer communities with high link density and composed by nodes having similar attributes. Experiments on synthetic networks show that the method finds communities almost equal to the ground-truth ones. Moreover, we compared our approach with a clustering method using only the attribute information, and a method which clusters nodes using only the multilayer network structure, on four real-world multilayer networks enriched with attributes. The results point out that the exploitation of the information coming from both all the layers and the node features allows the identification of accurate network divisions.