Research On Community Partitioning Algorithm Based On Similarity And Node Centrality

There are many complex systems in nature that can be described by a variety of networks.Features of complex networks include small worlds,scale-free,and community structures.In recent years,many concepts and methods of similarity have been successfully applied to the study of complex network structural features.The community structure describes the complex relationship between the network member nodes.The nodes within the community are closely connected,and the nodes between the community and the community are loosely connected.Deep understanding of the community structure in the complex network can better discover the hidden laws in the network.Most of the existing community partitioning algorithms have limitations in the complexity of the network scale,and the accuracy of some algorithms needs to be improved.This paper combines the similarity index with the node centrality index to propose a series of community partitioning algorithms,which reduces the computational complexity and improves the accuracy.The main work and results of this paper are as follows:1.A community partitioning algorithm based on local similarity of feature vectors is proposed.For the problem of high complexity of existing community partitioning algorithm,similarity index is used to combine node centrality.Based on the feature vector centrality of nodes in the network,feature vector locality is proposed.Similarity(ELS)and feature vector attraction(EA)indicators;ELS indicators represent similarities between nodes,used to form the initial community;EA indicators consider both local similarity and feature vector centrality ratio,indicating node The attraction is used to optimize the initial community,and on this basis,complete the community division of the network.2.A community partitioning algorithm based on PageRank local similarity is proposed.The problem that the central node of the high-node neighbor nodes from the feature vector is excessively differentiated due to the high number of nodes in the node is based on the PageRank centrality.PageRank local similarity(PLS)and PageRank Attraction(PA)indicators;PLS indicators indicate the degree of similarity between nodes,similar nodes aggregate into the initial community;PA indicators indicate the degree of attraction between nodes,used to optimize the community,realize the network Division of the community.3.A community partitioning algorithm based on structural similarity is proposed.According to the influence of the similarity of the connected nodes in the network on the edge weight,based on PageRank centrality,PageRank local dissimilarity(PLD)and PageRank enhanced similarity(PES)indicators are proposed.The PLD indicator indicates the degree of dissimilarity between nodes,which is used to assign values to the connected edges between nodes,delete the connected edges with higher PLD values,and disconnect the network to form the initial community partition;the PES index indicates the degree of similarity between the initial communities.To merge similar small community structures to complete the community division of the network.In this paper,the similarity algorithm is applied to the community division of complex network and combined with the central characteristics of the nodes in the network,a series of indicators are proposed to divide the community.