A new stochastic diffusion model for influence maximization in social networks

Most current studies on information diffusion in online social networks focus on the deterministic aspects of social networks. However, the behavioral parameters of online social networks are uncertain, unpredictable, and time-varying. Thus, deterministic graphs for modeling information diffusion in online social networks are too restrictive to solve most real network problems, such as influence maximization. Recently, stochastic graphs have been proposed as a graph model for social network applications where the weights associated with links in the stochastic graph are random variables. In this paper, we first propose a diffusion model based on a stochastic graph, in which influence probabilities associated with its links are unknown random variables. Then we develop an approach using the set of learning automata residing in the proposed diffusion model to estimate the influence probabilities by sampling from the links of the stochastic graph. Numerical simulations conducted on real and artificial stochastic networks demonstrate the effectiveness of the proposed stochastic diffusion model for influence maximization.