Research On Overlapping Community Detection Based On Center Edge Selection

Overlapping community structure is ubiquitous in social networks,information networks,technology networks and biological networks,and contains meaningful information.With the increasing complexity of community network structure,the potential value of network becomes more and more meaningful.More and more scholars have devoted themselves to the research of overlapping community discovery,which has promoted the overlapping community discovery algorithm and the evaluation method of overlapping community.Rapid development.The overlapping community discovery algorithm based on center node selection(CNS)is a traditional overlapping community discovery algorithm.The main content of CNS algorithm is the process of center node selection and clustering.In the process of center node selection,the central nodes in the network,i.e.the key nodes in the network,are selected according to the relationship between the nodes.In the clustering process,the central nodes are selected according to the central nodes.Points are used to divide other nodes in the network.However,in the CNS algorithm,the importance of the influence between nodes to the partition of community network is not considered,which reduces the accuracy of the CNS algorithm.Moreover,the CNS algorithm is based on the partition of nodes,and it is difficult to obtain a more appropriate overlapping node rate.To solve these two problems,this paper proposes an overlapping community discovery algorithm(CES)based on center-edge selection.There are three main improvements: in the process of center-edge selection,CES introduces the theory of community magnetic interference(CMI).In CMI,not only the information of the node itself,but also the interaction between the node and the node,each central node is considered.Point selection will affect the surrounding nodes,according to certain rules,the weight of nodes can be adjusted reasonably,so that the results of the center selection process are more accurate and more in line with the true distribution of the network;in the process of community division,we make full use of the relationship between edges and get the result of edge division,so that we can get the appropriate result after transforming it into the result of node division.Overlapping node rate can better discover the key nodes in the network,more fully understand the potential structure of the network,mining more meaningful information.In the process of overlapping node pruning,we optimize the adjustment of community network proposed by Wu,Z and others in 2010,expand the direction of pruning,prune overlapping nodes from different angles,and make the final community division.The result is closer to the actual network.The experimental results show that CES performs better than other three overlapping community discovery algorithms(CNS)in three standard community networks(Karate,Dolphin and Football)and three protein interaction networks(M.musculus,E.coli and Celevisiae)under three classical overlapping community evaluation indicators(EQ,NMI and CR).The faction filtering algorithm CPM and LC algorithm.