Research On Discovery Method Of Large-scale Complex Network Community With Fuzzy Structure

The study of community structure can not only deeply understand the complex network,but also tap the potential function of complex network.But with the development of information,the scale of complex network is more and more large.The flexibility of network makes the structure of complex network fuzzy,which makes it difficult for community discovery algorithm to achieve satisfactory results in dealing with large-scale complex network problems with fuzzy structure.In this dissertation,two improved methods are proposed to solve the problems of low computational efficiency of existing community discovery algorithms in large-scale complex networks and low accuracy of community discovery in complex networks with fuzzy structure.The main research contents are as follows:(1)In order to solve the problems of low computational accuracy and efficiency of existing community discovery algorithms in large-scale complex network community discovery,a multi-objective community discovery algorithm based on spectral clustering(SMOEA)is proposed.Firstly,the spectral clustering algorithm is used to deal with the encoded complex network,and the information of nodes and edges is fully utilized by the subgraph partition characteristics of spectral clustering to improve the quality of the initial population in the multi-objective community discovery algorithm.Secondly,the multi-objective community discovery algorithm is used to find the set of non-dominated solutions to obtain better solutions.In the evolutionary process of multi-objective particle swarm optimization,a data reduction method of grid reduction is used to reduce the population,improve the computational efficiency of the algorithm,and enable it to complete large-scale complex network community discovery problems.Experimental results on simulated networks and nine real networks show that the algorithm is superior to MRMOEA?RMOEA and MCMOEA in community discovery performance and computational complexity.(2)In order to solve the problem of performance degradation of existing community discovery algorithms in large-scale complex networks with fuzzy structure,a structure enhanced maximum group community discovery algorithm(MCSE)is proposed.Firstly,the algorithm increases the links between nodes that may belong to the same community,and removes the links between nodes in different communities,so that the community structure of complex network becomes clear.Secondly,we use the maximum group community discovery algorithm to deal with the complex network after structure enhancement,which is used to deal with the overlapping community discovery problem.Finally,by parallelizing the two parts of node similarity and search maximum clique set,the computing efficiency of MCSE algorithm is improved and it is convenient to deal with large-scale complex networks.Experimental results on simulated networks and four real networks show that the algorithm is superior to CPM,EdgeBoost and MCMOEA in community discovery performance and computational complexity.At the same time,the validity of parallel computing is proved by experiments.