Tracing influential nodes in a social network with competing information

Abstract

We consider the problem of competitive influence maximization where multiple pieces of information are spreading simultaneously in a social network. In this problem, we need to identify a small number of influential nodes as first adopters of our information so that the information can be spread to as many nodes as possible with competition against adversary information. We first propose a generalized model of competitive information diffusion by explicitly characterizing the preferences of nodes. Under this generalized model, we show that the influence spreading process is no longer submodular, which implies that the widely used greedy algorithm does not have performance guarantee. So we propose a simple yet effective heuristic algorithm by tracing the information back according to a properly designed random walk on the network, based on the postulation that all initially inactive nodes can be influenced by our information. Extensive experiments are conducted to evaluate the performance of our algorithm. The results show that our algorithm out-performs many other algorithms in most cases, and is very scalable due to its low running time. © Springer-Verlag 2013.