Complex Network Community Detection Based On Multiobjective Genetic Algorithm

Analysis of complex real-world networks has led to some significant discoveries in recent years. Studies have shown that community structure is prevalent in the real network topology from fields in nature, including various social, biological, Internet, economy, political and other networks. For instance, in social networks, individuals with common backgrounds or interests often found organizations together and communicate more frequently within the organizations than between them. Generally, a community, or the so called module in a network is considered as a node subset which has dense node-node connections internally but sparse connections externally. Revealing community structure of network is very important for analyzing function of network, discovering potential modes of network and forecasting action of network.At present, community detection is used to be transformed into optimization problem. And more and more swarm intelligence optimization algorithms have been used in community detection, but most of the algorithms optimized single objective. Although we achieved success on the theory and application of community detection based on single objective, but there are some problems, such as high complexity, the solution restrictions. In order to solve these problems, a natural method is to divide the community into as a multi-objective optimization problem.A multiobjective genetic algorithm to uncover community structure in complex network is studied.firstly, we constructed two objective functions,named community score and community fitness,which can identify inter-connections each other and have sparsely relationship.After returning non-dominated solution of a pair of objective functions by multiobjective genetic algorithm, the most suitable solution is selected through modularity and normalized mutual information of these two assessment indicators.This algorithm can discover hierarchical structure of network, consisting of a higher number of modules, are contained in solutions having a lower number of communities. The number of modules is automatically determined by the better tradeoff values of the objective functions. Finally, Experiments on synthetic and real life networks show that the algorithm successfully detects the network structure and it is competitive with state-of-the-art approaches. And the experiments also show that the algorithm can be effectively applied to large-scale complex network is divided in large networks.