PRDiscount: A Heuristic Scheme of Initial Seeds Selection for Diffusion Maximization in Social Networks

Abstract

This paper focused on seeking a new heuristic algorithm for the influence maximization problem in complex social network, that is, how to effectively select a subset of individuals to trigger a large cascade of further adoptions of new behavior under certain influence cascade models. In literature, degree and other centrality-based heuristics are commonly used to estimate the influential power of nodes in social networks. One of major issues with degree-based heuristics is that they are derived from the uniform IC (independent cascade) model, where propagation probabilities on all edges are the same, which is rarely the case in reality. Based on the general weighted cascade model (WC), this paper proposes Pagerank-like heuristic scheme, PRDiscount, in which discounting the influence power is explicitly adopted to alleviate the “overlapping effect” occurred in behavior diffusion. Then, we use both the artificially constructed social network graphs (with the features of power-law degree distribution and small-world characteristics) and the real-data traces of social networks to verify the performance of our proposal. Simulations illustrate that PRDiscount can advantage over the existing degree-based discount algorithm, DegreeDiscount, and achieve the comparable performance with greedy algorithm.