Research On Community Detection Algorithm Based On Complex Network Node Characteristics

With the rapid popularization of modern network world,data mining technology is becoming more and more mature.Under these conditions,complex networks have become a hot research topic in the field of data mining.One of the hot topics of complex network is community detection,and cluster analysis is the main technical means of its exploration.The analysis of community structure in complex networks can help researchers understand the structural performance and evolution of complex networks more deeply.In this paper,the network model of complex network,the basic concept of cluster analysis and the common community detection algorithm are discussed.After analyzing and studying the structure and performance of complex network,this paper proposes two community detection algorithms with better performance for the characteristics of complex network nodes,and applies them to graph clustering successfully.(1)A community detection algorithm based on node proximity-CS-Cluster is proposed.The algorithm is used for cluster detection of nodes with semantic information.Through research,it is found that most of the community detection algorithms in complex networks are proposed based on the research and analysis of node topology,and they ignore the semantic characteristics of nodes themselves,which leads to inaccurate community detection results.Based on this phenomenon,a new concept(node proximity)is proposed to complete the calculation of the phase dissimilarity between nodes.The concept of correlation degree and matching degree is introduced to calculate the structural dissimilarity between nodes.Then the selection rule of the initial cluster center point is redefined.This method avoids the disadvantage of artificial judgment and improves the accuracy of clustering.Finally,CS-Cluster algorithm uses the framework of K-Medoids algorithm to complete the community division.Experiments are carried out on two data sets,and the results show that CS-Cluster achieves good clustering effect.(2)An algorithm based on node core degree-NCD is proposed to detect overlapping communities.Firstly,the concept of node core degree is proposed.Secondly,based on the decision of whether the triangle model can be built between nodes in the network,the overlapping communities of the network is preliminarily detected by using the principle of center point extensibility.Finally,the overlapping nodes are identified on the basis of the divided communities.In this paper,NCD algorithm is applied to three standard data sets for experiments,and NMI and modularity are used to evaluate the experimental results.Experimental results show that NCD algorithm can effectively detect overlapping communities.