PSO-Based Community Detection in Complex Networks

Community detection is always an outstanding problem in the study of networked systems such as social networks and computer networks. In this paper, a novel method based on particle swarm optimization is proposed to detect community structures by optimizing network modularity. At the beginning, an improved spectral method is used to transform community detection into a cluster problem and the weighted distance which combine eigenvalues and eigenvectors is advanced to measure the dissimilarity of two nodes. Then, PSO is employed for cluster analysis. There are two definitive features in our algorithm: first, the number of communities can be determined automatically; second, the particle has low-dimensional structure by using only the corresponding components of the first nontrivial eigenvector to express community centers. The application in three real-world networks demonstrates that the algorithm obtains higher modularity over other methods (e.g., the Girvan-Newman algorithm and the Newman-fast algorithm) and achieves good partition results.