Complex Network Community Detection Based On Nodes Influence

Many systems in the real world can be described by complex networks such as social networks,citation networks in academic research,metabolic networks,protein networks,power networks,transport networks,scientists’ networks,the World Wide Web,the neural network,and so on.At present,the researchers of complex network focus on meso-scale structures and microstructure.The microstructure mainly analyzes the importance of nodes to reveal the network characteristics,and the mesoscopic structure focus on the clustering and community detection in the network.It is an important issue in the study of complex network by using quantitative analysis method to measure the importance of nodes in large-scale network.Different definitions of node importance caused different algorithms of the existing node evaluation method.There are local algorithms based on nodes’ location and neighbor,global algorithm based on eigenvector,and node removal shrinkage method for survivability test.The Eigenvector Central(EC)algorithm has a clear definition and high precision,and received great attention in academic and practical applications.However,the EC algorithm only considers the contribution of neighbor nodes and ignores other nodes,and also the loss of information during propagation.In view of the above analysis,we mainly improve the EC algorithm from the perspective of computing node influence,and put forward the community detection algorithm based on node influence.The main contributions of the paper are summarized as follows:(1)Proposed Accessbility Centrality(AC)algorithm.Firstly,the path calculation method based on topology is proposed,and the path of each node propagating to other nodes is obtained.The more of the paths,the greater the influence of nodes,which can locate the important nodes in the center of the topology quickly.The cumulative operation computes the infection range and accelerates the node convergence process.The algorithm is tested in four classic real Word networks,Email network,Protein network and Airport network to verify the feasibility of the algorithm.And compared with the traditional EC algorithm,LeaderRank algorithm and other four algorithms to test the performance of the new algorithm.(2)Proposed Rational LPA Based Node Influence(RLPBNI)algorithm.Firstly,by studying in detail the label propagation,we improve the update order of label propagation nodes according to the influence of node influence.Secondly,the rules of rational node label propagation are proposed.Finally,the community overlap is introduced into the community reintegration process to improve the accuracy of the algorithm.Comparing with LPA algorithm and other traditional algorithms in the real network and artificial network,the results show that our method obtains the better performance.(3)Applying the AC and RLPBNI algorithms to the LinkedIn social network,to test our algorithms applied in the social network practice of real life,and then evaluates and analyzes the results.