Identifying influential spreaders in complex networks based on entropy weight method and gravity law**Project support by the National Key Research and Development Program of China (Grant No. 2018YFF0301000) and the National Natural Science Foundation of China (Grant Nos. 71673161 and 71790613).

Abstract

In complex networks, identifying influential spreader is of great significance for improving the reliability of networks and ensuring the safe and effective operation of networks. Nowadays, it is widely used in power networks, aviation networks, computer networks, and social networks, and so on. Traditional centrality methods mainly include degree centrality, closeness centrality, betweenness centrality, eigenvector centrality, k-shell, etc. However, single centrality method is one-sided and inaccurate, and sometimes many nodes have the same centrality value, namely the same ranking result, which makes it difficult to distinguish between nodes. According to several classical methods of identifying influential nodes, in this paper we propose a novel method that is more full-scaled and universally applicable. Taken into account in this method are several aspects of node’s properties, including local topological characteristics, central location of nodes, propagation characteristics, and properties of neighbor nodes. In view of the idea of the multi-attribute decision-making, we regard the basic centrality method as node’s attribute and use the entropy weight method to weigh different attributes, and obtain node’s combined centrality. Then, the combined centrality is applied to the gravity law to comprehensively identify influential nodes in networks. Finally, the classical susceptible-infected-recovered (SIR) model is used to simulate the epidemic spreading in six real-society networks. Our proposed method not only considers the four topological properties of nodes, but also emphasizes the influence of neighbor nodes from the aspect of gravity. It is proved that the new method can effectively overcome the disadvantages of single centrality method and increase the accuracy of identifying influential nodes, which is of great significance for monitoring and controlling the complex networks.