An Improved Evaluating Method of Node Spreading Influence in Complex Network Based on Information Spreading Probability
Ruan Yi-Run,Lao Song-Yang,Wang Jun-De,Bai Liang,Hou Lu-Lin
DOI: https://doi.org/10.7498/aps.66.208901
2017-01-01
Abstract:How to evaluate the node spreading ability and how to find influential nodes in complex networks are crucial to controlling diseases and rumors, accelerating or hindering information from diffusing, and designing effective advertising strategies for viral marketing, etc. At present, many indicators based on the shortest path, such as closeness centrality, betweenness centrality and the (SP) index have been proposed to evaluate node spreading influence. The shortest path indicates that the information transmission path between nodes always selects the optimal mode. However, information does not know the ideal route from one place to another. The message does not flow only along geodesic paths in most networks, and information transmission path may be any reachable path between nodes. In the network with high clustering coefficient, the local high clustering of the nodes is beneficial to the large-scale dissemination of information. If only the information is transmitted according to the optimal propagation mode, which is the shortest path propagation, the ability to disseminate the node information would be underestimated, and thus the sorting precision of node spreading influence is reduced. By taking into account the transmission rate and the reachable path between a node and its three-step inner neighbors, we design an improved method named ASP to generate ranking list to evaluate the node spreading ability. We make use of the susceptible-infected-recovered (SIR) spreading model with tunable transmission rate to check the effectiveness of the proposed method on six real-world networks and three artificial networks generated by the Lancichinetii-Fortunato-Radicchi (LFR) benchmark model. In the real data sets, the proposed algorithm can achieve a better result than other metrics in a wide range of transmission rate, especially in networks with high clustering coefficients. The experimental results of the three LFR benchmark datasets show that the relative accuracy of ranking result of the ASP index and the SP index changes with the sparseness of the network and the information transmission rate. When the information dissemination rate is small, SP index is slightly better than the ASP index. The reason for this result is that when the transmission rate is small, the node influence is close to the degree. However, when the transmission rate is greater, the accuracy of the ASP index is higher than those of other indicators. This work can shed light on how the local clustering exerts an influence on the node propagation.