Multiple evidence fusion based information diffusion model for social network

Study on diffusion behaviors, such as contagion of virus, adoption of production, and information propagation, is one of the hottest topics of social network research. The conventional models usually focus on a particular factor that influences the contagion process. A novel model of diffusion based on multiple behavior evidence fusion is proposed on the analysis of prevalent social network data set, which can simulate the information propagation process with the fitting goodness of 0.79. In this model, impact factors, i.e. the influence between neighborhood, and the activity of user, are taken into account together and fused into combined evidence mass function, which can be used to infer users´ forwarding behavior by D-S theory reasoning. Series of experiments tested on the published Enron email data set show that the multiple evidence fusion based diffusion model (MEFDM) is reasonable and feasible in demonstrating the cascading of information in the social network.