Time is What Prevents Everything from Happening at Once: Propagation Time-conscious Influence Maximization (1705.10977v2)

Abstract: The influence maximization (IM) problem as defined in the seminal paper by Kempe et al. has received widespread attention from various research communities, leading to the design of a wide variety of solutions. Unfortunately, this classical IM problem ignores the fact that time taken for influence propagation to reach the largest scope can be significant in realworld social networks, during which the underlying network itself may have evolved. This phenomenon may have considerable adverse impact on the quality of selected seeds and as a result all existing techniques that use this classical definition as their building block generate seeds with suboptimal influence spread. In this paper, we revisit the classical IM problem and propose a more realistic version called PROTEUS-IM (Propagation Time conscious Influence Maximization) to replace it by addressing the aforementioned limitation. Specifically, as influence propagation may take time, we assume that the underlying social network may evolve during influence propagation. Consequently, PROTEUSIM aims to select seeds in the current network to maximize influence spread in the future instance of the network at the end of influence propagation process without assuming complete topological knowledge of the future network. We propose a greedy and a Reverse Reachable (RR) set-based algorithms called PROTEUS-GENIE and PROTEUS-SEER, respectively, to address this problem. Our algorithms utilize the state-of-the-art Forest Fire Model for modeling network evolution during influence propagation to find superior quality seeds. Experimental study on real and synthetic social networks shows that our proposed algorithms consistently outperform state-of-the-art classical IM algorithms with respect to seed set quality.