Timing Matters: Influence Maximization in Social Networks Through Scheduled Seeding

One highly studied topic in the field of social networks is the search for influential nodes that, when seeded (i.e., activated intentionally), may further activate a large portion of the network through a viral contagion process. Indeed, various mathematical models were proposed in the literature to characterize the dynamics of such diffusion processes, and different solutions were suggested for maximizing influence under such models. However, most of these solutions focused on selecting a set of nodes to be seeded at the initial phase of the diffusion process. This paper suggests a scheduled seeding approach that aims at finding not only the best set of nodes to be seeded but also the right timing to perform these seedings. More specifically, we identify three different properties of existing contagion models that can be utilized by a scheduled approach to improve the total number of activated nodes: 1) stochastic dynamics; 2) diminishing social effect; and 3) state-dependent seeding. By analyzing each of these properties separately, we demonstrate the advantages of the scheduled seeding approach over the traditional initial seeding approach, both by theoretical and empirical evaluation. Our analysis presents an improvement of 10%-70% in the final number of infected nodes when using the scheduled seeding approach. Our findings have the potential to open up a new area of research, focusing on finding the right timing for seeding actions, thereby helping both in improving our understanding of information diffusion dynamics and in devising better strategies for influence maximization.