Influence maximization in social networks: An ising-model-based approach

The past few years have seen increasing interest in understanding social networks as a medium for community interaction. A major challenge has been to understand various fundamental properties of social networks that form the basis for the formation and propagation of opinions across such networks. The main hurdle has been the absence of plausible models that specify the correlations between different members of a social network, which could then be used for algorithm design. This paper studies an influence maximization problem using an Ising-model-based approach. We first validate the credibility of the ferromagnetic Ising model in predicting opinion formation in social networks using cosponsorship data from the US Senate proceedings. We then develop a greedy placement algorithm that can efficiently find an appropriate subset of network members, “bribing” whom can efficiently propagate a particular opinion in the network. We use simulations to confirm the effectiveness of the greedy placement algorithm.