A Method Towards Community Detection Based on Estimation of Distribution Algorithm

Abstract

Estimation of Distribution Algorithm (EDA) is a stochastic optimization algorithm based on statistical theory. It has strong global search ability, but it is easy to fall into the local optimal solution and can not get good results in community detection. In order to solve this problem, we propose a community detection algorithm based on Estimation of Distribution Algorithm, named EDACD, whose basic framework refers EDA and the target function is modularity. EDACD keeps population diversity by adding crossover mutation operation of Genetic Algorithm as well as the improvement of probability model. Genetic Algorithm is based on “micro” level of gene, which has good local optimization ability; EDA uses the evolutionary method based on “macro” level of search space, which has strong global search ability and fast convergence speed. Taking advantage of the two methods, EDACD can used to improve the search ability of algorithm from “micro” and “macro” two levels. Finally, by experimenting on some typical real-world networks and computer-generated networks, the experimental results show that the proposed algorithm can detect the community division accurately, and has higher clustering precision compared with some representative algorithms. In addition, the proposed algorithm also has a fast convergence rate.