Algorithm For Detecting Community Of Complex Network Based On Clustering

Complex networks is a kind of models and tools used in the study of complex system.As the complex networks of academic research,people refine a lot of nature of the complex networks.With the nature,community structure is a topological properties of complex networks,which is the most important one.If we can find that the community structure of complex networks,which is can be used in the analysis of the complex networks topology structure,function and implicit model etc.,and can fully predict various behavior of complex network,so the implementation of this behavior in the World-Wide-Web,social networks and biological networks is widely applicable.In this paper,the achievements of community detection research at home and abroad and analysis of relevant theories were expounded,and the advantages and disadvantages of the traditional community detection algorithms are summarized.Aiming at these problems,such as inefficiency and priori conditions,a spectrum algorithm based on MVV is proposed to improve the accuracy and efficiency.First,a method is found for nodes in the network transforming into the vectors—MVV method,and with which nodes can be changed into the data structure for clustering algorithm;Second,after analyzing the influence of the similarity measure on the effect of the clustering algorithm,a measure of diversity index based on Euclidean distance is proposed;Finally,aiming at the problem of hierarchical clustering,which do not apply to the large-scale networks because of noise,the paper combine spectrum clustering and hierarchical clustering.Spectrum clustering is used to reduce dimension of the vectors,and then hierarchical clustering based on diversity index is carried out.This paper use Matlab for data processing,to verify the spectrum algorithm based on MVV in aspects of accuracy,modularity,efficiency and large-scale networks.The results of experiment show that the algorithm can not only guarantee the quality of divided community in true networks and computer generated networks,but also improve the efficiency.And it is applicable to large-scale networks.In the end of this paper,the direction of further study in community detection is proposed.