A Research On Node Centrality Technology Based On Network Topology

With the development of the large-scale and dynamic Internet,network topology,as the prerequisite and foundation of network management,makes an essential contribution to improve network performance and ensure network stability.As an important offshoot of network topology management,node centrality has an extensive range of applications in network topology visualization,network classification and network reconstruction.However,with the current wide-ranging and scalefree Internet,there is a lack of node centrality metrics that can be effective for multiple Internet structures.According the corcern of current node centrality research,a new node centrality with the utilization of network subgraph,called Local Subgraph Centrality,is proposed.Local Subgraph Centrality is proposed and improved in undirected networks,directed networks and directed weighted networks.Furthermore,Local Subgraph Centrality is compared with other five basic node centralities from two aspects,which are the consistency with SIR simulation and the distinction of nodes.Finally,the effectiveness and superiority of Local Subgraph Centrality is verified by the simulation and experiment.1.In the undirected network,the derivation and definition of the Local Subgraph Centrality is proposed by combining the concept of subgraph.At first,based on the Three Degrees of Influence Rule,the second-level neighbors are divided into insular triangular nodes and diffuse star nodes by introducing triangular and star subgraph.Then,the definition of the Local Subgraph Degree is raised according to the classified second-level neighbor nodes.In addition,the Local Subgraph Centrality of nodes is derived from the Local Subgraph Degree.Finally,the simulation and comparison experiment are carried out in the Karate network with 34 nodes and the autonomous system network with 3570 nodes respectively.In the Karate network,the consistency and distinction of the Local Subgraph Centrality are 0.900 and 0.853,respectively,which are significantly improved compared to the suboptimal close centrality of 0.703 and 0.588.In the autonomous system network,the consistency and distinction of Local Subgraph Centrality are 0.744 and 0.419,respectively.The consistency is higher than the suboptimal close centrality of 0.676,and the distinction is lower than the close centrality of 0.507.2.In the directed network,the Local Subgraph Centrality is expanded to divergent Local Subgraph Centrality and aggregated Local Subgraph Centrality according to the directionality of the network.At first,in accordance with the directionality of the subgraph,the second-level neighbors are expanded into four types: aggregated triangular nodes,divergent triangular nodes,aggregated star nodes and divergent star nodes.Then,the Local Subgraph Degree is extended to divergent and aggregated Local Subgraph Degree based on the expanded secondary neighbors.In addition,pursuant to the Local Subgraph Degree,the Local Subgraph Centrality is distended to divergent and aggregated Local Subgraph Centrality.Finally,the simulation and experiment are carried out in a decentralized unstructured P2 P network with 6301 nodes.In the P2 P network,the consistency and distinction of the Local Subgraph Centrality are 0.951 and 0.094,respectively.Among them,the consistency is higher than the suboptimal H-index of 0.904,and the distinction ranks third,which is lower than the close centrality of 0.750 and the betweenness centrality of 0.373.The low distinction is mainly caused by the symmetry of the distributed unstructured P2 P network.3.In the directed weighted network,the divergent Local Subgraph Centrality and the aggregated Local Subgraph Centrality are expanded by introducing edge weights.At first,the coefficient related to edge weight is added to the divergence and aggregation Local Subgraph Degree.Then,according to the corresponding Local Subgraph Degree,the divergence and aggregated Local Subgraph Centrality in the directed weighted network are proposed.Finally,simulation and experimental verification are carried out in four directional weighted hyperlink networks.The number of nodes in these four networks is 349,286,433,and 300,respectively.In the hyperlink network,the consistencies of the Local Subgraph Centrality are 0.764,0.810,0.826 and 0.767,which are improved compared to the suboptimal close centrality of 0.659,0.676,0.729 and 0.677;and the distinctions are0.642,0.694,0.575 and 0.703,respectively.Compared with the close centrality of 0.622,0.608,0.612 and 0.673,the distinction of three networks is better than close centrality.