A Genetic Algorithm for Community Detection in Attributed Graphs

Abstract

A genetic algorithm for detecting a community structure in attributed graphs is proposed. The method optimizes a fitness function that combines node similarity and structural connectivity. The communities obtained by the method are composed by nodes having both similar attributes and high link density. Experiments on synthetic networks and a comparison with five state-of-the-art methods show that the genetic approach is very competitive and obtains network divisions more accurate than those obtained by the considered methods.