Research On Community Detection Algorithm Based On Lévy Flight

Most complex systems in the real world,such as mobile communication,transportation,research and education,social relations and so on,could be abstracted as complex networks.As one of the most important characteristics of complex networks,community structure contains a lot of valuable information.Discovering the community structure in a complex network can help us understand the structure of the real network better.It is of great significance in predicting the function of the network as well.The existing community detection methods can be roughly divided into three categories based on the classification results: non-overlapping community detection algorithms,overlapping community detection algorithms and dynamic community detection algorithms.To obtain a community structure with high accuracy and high quality,this paper focuses on non-overlapping and overlapping community detection algorithms.The main contributions are as follows:On one hand,in traditional multi-objective cuckoo community detection algorithm,the location of the nest is updated mainly based on Lévy flight,which leads to a slower convergence rate in the late search period.Therefore,this paper proposes a multi-objective community detection algorithm based on Lévy flight and discrete differential evolution strategy(MOCL-DDE).Multi-objective cuckoo search algorithm is used as a framework.Objective functions NRA and RC are then optimized to obtain a reasonable-scale community structure.To meet the demand of discrete community detection,the cosine function is introduced as a mapping operator.When updating the location of nest,Lévy flight and discrete differential evolution algorithm are combined in order to increase the population diversity and improve the convergence speed.On the other hand,to overcome the drawback that most intelligent optimization methods often fall into the local optimum in solving the overlapping community division,an algorithm based on Lévy flight(LFOCDA)is applied into overlapping community detection.This method combines the unique advantages of Lévy flight in the cuckoo algorithm.Extended modularity function is taken as a quality metric to obtain overlapping communities.In the iterative update period,variable-step random walk characteristics of Lévy flight give the algorithm an opportunity to jump out of the local optimum and expand the search scope,so that it can make one step towards the global optimal solution.