Application Research Of Spectral Clustering Method In Community Detection

There is a certain community structure in the complex network.The community structure reflects the degree of aggregation of nodes in the network,that is,the community is divided by the degree of aggregation of nodes.In recent years,with the rapid development of complex network research,detecting community structures in complex networks can better discover and analyze network characteristics.Among the many algorithms for detecting community structure,the spectral partitioning method transforms the process of maximizing modularity into the spectral problem of the matrix,which reduces the complexity of the algorithm and obtains better segmentation results.On this basis,this paper proposes a spectral clustering method based on edge clustering coefficients to divide the community and conduct experiments in real network data.The main contents of this article are as follows:(1)This paper introduces the background,significance,basic concepts and spectral clustering of complex networks.The method of graph partition based on Laplace matrix is analyzed and the graph partition problem is transformed into the solution problem of eigenvalues and eigenvectors in Laplace matrix.(2)Based on the distribution of the nodes in a complex network community characteristics were analyzed,and the internal connection between nodes is known in the community is more dense,and node connection between community is often sparse,the network clustering coefficient is good interpretation of this characteristic,the edge of the clustering coefficient reflects the node connection relationship between adjacent nodes.Based on this,proposed based on clustering coefficient of spectral clustering community classification method: based on the network within the community in the edge between the community was more triangles contained,defines the clustering coefficient matrix and the gain function,and the matrix eigenvalue and eigenvector of the division,so as to achieve the community,in the end,through experiments in real network data,verify the feasibility of this method.