Research On Community Detection Algorithm Based On Node Importance And Similarity

Community detection is a technology that uses graph topology information to parse community structure from complex networks.It can be widely used in social networks,biomedicine and other fields.Community detection algorithms are mainly divided into graph partition algorithm,modularity optimization algorithm,label propagation algorithm,similarity algorithm,etc.Among them,label propagation algorithm has been widely concerned with other types of algorithms due to its nearly linear time complexity and excellent community detection quality.However,the algorithm has the problem of high randomness in the propagation process and can’t obtain stable community partition.Therefore,considering the importance and similarity of nodes,this dissertation adds them to the label propagation process,and proposes two improved community detection algorithms.The specific work contents are as follows:(1)Community detection algorithm based on node importance(Community Detection Algorithm Based On Node Importance)is proposed.Firstly,the algorithm completes the initial division of the community by assigning labels to the seed nodes and their neighbor nodes;Secondly,the communities are sorted according to the number of nodes,and the node label change is introduced into the community to redefine the node importance,so as to determine the node update order;Then,when multiple maximum labels appear as candidates,the direction of label propagation is determined according to the node influence defined by the joint action of node importance and node superiority;Finally,the value of modularity before and after the community is used as the basis to determine whether the community is merged.However,the seed node selection of the algorithm is relatively simple,the importance of this paper leads to the first propagation of important nodes,which limits the propagation range of nodes.(2)Community detection algorithm based on node similarity(Community Detection Algorithm Based On Node Similarity)is proposed.To solve the problems of the above algorithms,the algorithm in this chapter first uses the similarity of seed nodes to select seed nodes to complete the initial community division in the initial community division stage;Secondly,the similarity between nodes and neighbor nodes is introduced in the community to optimize the importance of nodes and complete the ranking of label propagation in the community;Then,when there are multiple candidate tags,the reverse influence of neighbor nodes is considered and added to the node influence calculation,so as to obtain the accurate update order;Finally,in order to improve the quality of the community,the community is merged according to the community similarity,so as to promote the algorithm to detect and obtain high-quality community division.In this dissertation,comparative experiment is carried out on artificial network and real network to verify that the two algorithms have achieved better experimental results in terms of stability and accuracy.The community detection results show that the two algorithms proposed in this dissertation can effectively detect the community structure.Figure [20] Table [6] Reference[68]...