The comparison of four methods in finding influential spreader in social network
暂无分享,去创建一个
With the prevalence of smart phone and Internet, social media have been developing both its customer base and its impact on the society. With the use of social media's influence on people, companies begin to advertise their product and services on the social media platform. Therefore, in order to guarantee the effect of the advertisement, methods to find influential spreaders in social network are proposed. In this paper, we aim to compare four different methods that have been proved to be efficient, on both simulated network and real life network, and then use SIR model to propagate their diffusion process. The results of simulation network show that all four methods produced the same influential spreaders and it was proven to be a good initial spreader in a SIR model. However, in the network generated from Sina Weibo, the four methods all produced different influential spreader. The possible cause of the different results from simulated result is that the simulated network neglects the cohesion of influential nodes and therefore, the simulated network will value degree more than the position of a node. After applying the SIR model, we found that the degree and betweenness produce nodes that have the highest infection number while K-core decomposition produces nodes that can reach nodes in a further distance faster than other methods and nodes produced by PageRank can reach farther in the network despite of its less infection number.
[1] Scott Counts,et al. Identifying topical authorities in microblogs , 2011, WSDM '11.
[2] M. Barthelemy. Betweenness centrality in large complex networks , 2003, cond-mat/0309436.
[3] Lev Muchnik,et al. Identifying influential spreaders in complex networks , 2010, 1001.5285.
[4] Matteo Pontecorvi,et al. Betweenness Centrality - Incremental and Faster , 2013, MFCS.