Research On Community Discovery Method Based On Importance Of Nodes Links In Complex Network

Community discovery helps to discover the network structure and dynamics characteristics,and then has a positive impact on people's life.It is mainly manifested in real-time public opinion monitoring,system intelligent recommendation,traffic safety detection,stable power transmission,etc.Therefore,research on community discovery has important meaning.This article mainly includes the following three parts:(1)In the GN community discovery algorithm,there are problems such as not considering the influence of the path length when calculating the edge betweenness centrality and the difference in the amount of accumulated information under different path lengths.In this paper,considering the influence of the long paths and edge weights of different path lengths,a community discovery algorithm WLCD based on edge weights and local edge betweenness centrality is proposed.First,when the algorithm is initialized,all nodes are set in the same community,and calculate the weighted local edge betweenness centrality of each edge in the network.Secondly,choose to remove the edge with the largest weighted local edge betweenness centrality,and iteratively loop until there is no edge in the network.Finally,the optimal modularity is used as the objective function to output the community division result.(2)Asynchronous label propagation community discovery algorithm in the propagation process,the random selection of neighbor node label update and node update sequence is not clear,resulting in poor accuracy and convergence of algorithm division results,and it is easy to form a giant community.Aiming at random situations such as label selection and node update order,considering the influence of edges and nodes,this paper proposes a stable asynchronous label propagation community algorithm SALPA based on the influence of edges and nodes.First,the algorithm initialization sets a unique label for each node,and takes the direct and indirect contributions of the nodes as edge influence.Secondly,calculate the influence of network nodes based on edge influence and feature vector,and rank the influence of nodes.Finally,the node influence sequence is used as the update sequence,and the label weight is used to guide the asynchronous update process of the node label.Then nodes with the same label are divided into the same community and the community division result is output.(3)The semi-synchronous label propagation algorithm initially sets the label without considering the network substructure,and there is randomness in label selection during the update process,resulting in poor algorithm accuracy and convergence.This article considers the impact of substructure and label importance,propose a semi-synchronous label propagation community discovery KSLPA algorithm based on K-cliques and label importance.First,the set of K-cliques in the network is used as a substructure,and the node label is set to the largest node ID number in the set.Secondly,the nodes in the network are colored using the greedy idea,so that the colors between any two adjacent nodes are different.Finally,the importance of the node label is calculated,and the semi-synchronous update process is jointly guided by the cumulative value of the same label importance and the node color sequence.Then the nodes with the same label are divided into the same community and the community division result is output.Finally,on real networks such as Karate,Dolphins,Polbooks and Football networks,artificial reference networks under different parameters,and artificially crawled Wikipedia networks,the algorithm of this article is combined with GN,FN,LPA,Louvain,Spinglass,asynchronous label propagation,synchronous label propagation,Semi-synchronous label propagation,LPAm,CNM algorithms are compared with experiments to verify the effectiveness and accuracy of the algorithms in this paper.