Self-adaptive Differential Evolution for Community Detection

Abstract

The detection of community structure in complex networks is an important problem deeply investigated in the last years. In fact, the awareness of network organization allows a better understanding of network properties, which could not be captured when studying the network as a whole. Evolutionary computation techniques, including Genetic Algorithms (GAs) and, more recently, Differential Evolution (DE), showed to be competitive techniques for the solution of this problem. In this paper, a new method for community detection based on DE is proposed. The approach employs different mutation and crossover operators, which are chosen at random at each iteration. Moreover, it introduces a self-adaptive strategy that changes part of the population and the scaling factor when the fitness function does not improve for a number of generations. Experiments on real-world and synthetic networks show that the method obtains good performance and it is competitive with respect to other DE-based algorithms.