Influence maximization in social media networks concerning dynamic user behaviors via reinforcement learning

Abstract This study examines the influence maximization (IM) problem via information cascades within random graphs, topology of which dynamically changes due to uncertainty user behavior. leverages discrete choice model (DCM) calculate probabilities existence directed arc between any two nodes. In this IM problem, DCM provides a good description and prediction behavior in terms following or not neighboring user. To find maximal at end finite-time horizon, models by using multistage stochastic programming, can help decision-maker select optimal seed nodes broadcast messages efficiently. Since computational complexity grows exponentially with network size time original is solvable reasonable time. then uses different approaches approximate decision: myopic two-stage programming reinforcement learning Markov decision process. Computational experiments show that method outperforms method.