Maximizing information or influence spread using flow authority model in social networks

Abstract

Identifying a set of nodes of size k in a large social network graph which maximizes the information flow or influence spread is a classical subset selection problem which is NP-Hard. Recently Charu Agarwal et al. in paper [10] proposed a stochastic information flow model and two algorithms namely, RankedReplace and BayesTraceBack to retrieve influential nodes. Among the two, RankedReplace algorithm gives better information spread, but does not scale well for large data sets. The main objective of paper is to speed up the RankedReplace algorithm without compromising on the information spread. To achieve this we are using the idea of degree discount heuristic from [8] and maximum degree heuristic. As shown by our experimental results, the proposed modifications reduce the amount of time significantly, maintaining the influence spread almost equal and even marginally better at times as compared to RankReplace algorithm. We have also proposed Willingness to send heuristic(WS) and an algorithm based on this WSRank for directed social network graphs. © 2014 Springer International Publishing Switzerland.