Adaptive community detection in complex networks using genetic algorithms

Community detection is a challenging optimisation problem that consists in searching for communities that belong to a network or graph under the assumption that the nodes of the same community share properties that enable the detection of new characteristics or functional relationships in the network. A large number of methods have been proposed to address this problem in many research fields, such as power systems, biology, sociology or physics. Many of those optimisation methods use modularity to identify the optimal network subdivision. This paper presents a new generational genetic algorithm (GGA+) that includes efficient initialisation methods and search operators under the guidance of modularity. Further, this approach enables a flexible and adaptive analysis of the characteristics of a network from different levels of detail according to an analyst's needs. Results obtained in networks of different sizes and characteristics show the good performance of GGA+ in comparison with other five genetic algorithms, including efficient algorithms published in recent years.