Hyper-IMRANK: Ranking-based Influence Maximization for Hypergraphs

Influence maximization algorithms have found tremendous success in applications such as social media viral marketing. However, in several applications, the nature of relationships among participating entities cannot be accurately described only using pairwise interactions. To model group interactions, we consider the influence maximization problem in hypergraphs. Specifically, we consider a recently proposed diffusion model - HyperCascade which generalizes the standard independent cascade model. We propose a novel ranking based algorithm called Hyper-IMRANK to select highly influential nodes under this model. We demonstrate the superior performance of our algorithms in real world as well as synthetic datasets.