Fast PSO algorithm for community detection in graph

Communities are strongly connected structures in large networks. Community detection is of great importance in many disciplines where systems are often represented as graphs. In this paper, a fast method based on particle swarm optimization (PSO) is proposed for community detection. It is a divisive method and uses modularity Q as the fitness function of PSO algorithm. In the algorithm, a special encoding scheme based on the partition solution of a network is designed to represent the community partition. In order to reassign the isolated nodes to the neighbor community, movement strategy is used to make the partition result reasonable and improve the modularity value. The experiments on the real networks demonstrate that the algorithm can obtain high modularity value and achieves good community results.