Research On Community Detection Algorithm Of Spectral Clustering Based On Hierarchical Division

With the rapid development of Internet technology,complex network research has spread to all aspects of human society.Community detection,as a popular field in complex network research,provides an important theoretical basis for understanding network structure,functional characteristics of network and solving network social problems.In recent years,experts have also proposed a large number of excellent community detection algorithms from different perspectives,but there are also problems of different degrees.Based on in-depth study of the existing community detection algorithms,this paper proposes the following two algorithms combined with the hierarchical structure characteristics of the network:Due to the problems of the spectral clustering algorithm in the unweighted network,such as the number of input clusters,sensitive selection of initial cluster points and inaccurate community division results,an eH-SC algorithm for spectral clustering unauthorized network community detection algorithm based on average vector hierarchical optimization is proposed.Firstly,a density weighting function is proposed to measure the influence of different nontrivial feature vectors of network matrix on community aggregation.Then combined with the density weighted function defines the average vector containing the node properties and attributes of the network topology,again based on feature distance between nodes minimum level of average vector optimization for corporate merger.Finally combining with modularity function to detect the network community structure.For the weighted network,the current weighted network community detection algorithm mainly considers the edge weights of the network,but fails to consider the influence of nodes in the weighted network.In the spectral clustering algorithm,weight can be easily introduced into the network matrix,based on this,this chapter improves the eh-SC algorithm,and puts forward the spectral clustering weighted network community detection algorithm weh-SC algorithm based on the weighted feature distance.Firstly,the normal weighted matrix of the network is constructed by introducing the weight information,compute initial vector with the weight attribute,which has the global attribute of nodes in the network and the weight attribute of edges.Then the weighted feature distance is proposed to judge the similarity of nodes in the weighted network.Finally,the community structure of the weighted network is detected by hierarchical optimization of the initial vector and weighted modularity.In order to verify the effectiveness of the proposed algorithm,experiments were carried out on synthetic data sets and real social network data sets.Taking modularity and standardized mutual information as evaluation indexes.Compared with other algorithms,the results show that the two algorithms can classify high-quality community structures.