Research On Complex Network Community Detection Method Based On Node Similarity

A complex network is an abstract representation of a complex system,which reflects the connection relationship between individuals in the complex system.As the basic function of complex networks,communities generally exist in actual networks.By mining community structure,it is helpful to understand network functions and study network topology,so as to reveal the underlying laws in complex networks and predict the evolution trend of complex networks.Therefore,it has important practical application value for the research of complex network community detection.Experts started from various angles and proposed a wide variety of community detection algorithms.Among such algorithms based on node similarity community detection,the essence is to measure the similarity of nodes,but there are such algorithms based on node similarity.All have defects of varying degrees.In this paper,the following two algorithms are designed by improving the measurement of node similarity and the framework of fusion clustering algorithm.(1)Aiming at the problems that traditionally rely on node neighbors to measure the similarity of nodes,the accuracy of node similarity is low,and the random selection of clustering center nodes leads to the unstable quality of the final community detection.A community detection algorithm based on node density and similarity is proposed.The algorithm takes into account the influence of the shortest path of nodes on the similarity of nodes,and gives a method to improve the similarity of nodes by combining the neighbors of nodes and the shortest paths of nodes;defines the density of nodes to provide a basis for the selection of cluster center nodes.First,select the initial cluster nodes according to the principle of maximum node density and small similarity between nodes,then use the principle of maximum similarity to divide the community,and finally use the optimized modularity as an indicator to perform community detection.(2)In order to solve the problems of traditionally relying on node edge weights to measure the accuracy of node similarity and single community evaluation index,a community detection algorithm based on node similarity is proposed.The algorithm calculates the structural similarity of the nodes based on the attractiveness of the nodes,calculates the attribute similarity of the nodes based on the simple matching of the K pattern,and calculates the similarity of the nodes by combining the structural similarity and the attribute similarity of the nodes;the community closeness is introduced as a community Evaluation index at the time of merger.First,select the cluster center according to the principle of maximum node density,and then divide the communities based on node similarity.Finally,merge the communities according to the optimal community closeness until the closeness of the community no longer increases.At this time,the community detection result is obtained.In order to verify the performance of the above two algorithms,this paper has conducted sufficient experiments in artificial networks and real-world networks,and made an in-depth comparison between the experimental results obtained and other algorithms.The experimental results show that the two mentioned in this paper The algorithms can effectively detect high-quality community structures.