关键词: 社区发现, 重要节点发现, 节点影响力, 拓扑结构

Abstract: Identifying influential nodes in complex networks has been widely applied in reality. Conventional methods consist of local metrics and global metrics, and Eigenvector Centrality(EC)algorithm is the most representative method in global metrics. EC algorithm regards all nodes in network as one community and calculates the influence of nodes by considering the node influence of their neighbors. However, EC algorithm ignores the topology structure of community networks in reality, where nodes’ neighbors may come from different communities with different clustering coefficient. To address this issue, this paper proposes an Accessibility Centrality(AC)model based on the network topology. Firstly, the adjacency matrix is considered as the feedback path, and nodeinfluence is computed through different pathswith different length. Then, node influence from every step is revised, due to existing noise in the paths. At last, every nodes’ influence (AC value)is calculated via the revised result in the previous procedure. To evaluate the performance of our method, Susceptible-Infected-Recovered(SIR)model and Susceptible-Infected(SI)model are employed to simulate the process of influence propagation on four real networks. Comparing with Eigenvector Centrality, Betweenness Centrality, Degree Centrality and LeaderRank algorithms, the results show that our method obtains the highest accuracy.

Key words: community detection, influential nodes discovery, nodeinfluence, network topologic