PSO combined with LPA for the detection of overlapping community

Since community detection become a hot subject, lots of algorithms for overlapping community detection have been proposed. These algorithms may have some unstable factors which would bring some undesirable results. In order to improve the performance of the algorithm, a novel algorithm, PSO_LPA, based both on particle swarm optimization (PSO) algorithm and label propagation algorithm (LPA), is proposed. The PSO_LPA algorithm uses PSO algorithm as the framework to find the non-overlapping community structure. Based on the obtained non-overlapping community structure, the PSO_LPA algorithm marks each node with a label which describes the probability of the node belongs to the communities. With the help of the labels, the PSO_LPA algorithm can find the overlapping nodes by some weighted operations. The proposed algorithm is tested on several real-world networks. According to the experiment, the new algorithm, compared to the existing algorithms, is more efficient and accurate.