A PDE-based Model for Information Diffusion in Social Networks

Abstract

Online social networks (OSNs), as a novel and effective channel, play an important role in sharing and spreading information. A great deal of research investigates information diffusion process along time dimension, while relatively little effort has been made on the spatial and temporal dynamics. In this paper, a Recursive Diffusive (RD) Model based on Partial Differential Equation (PDE) is proposed to describe temporal-spatial trend of information diffusion, which reflects the effects of the network topology on information propagation. Based on the data crawled from twitter, parameters of the model are estimated by Levenberg-Marquardt algorithm, where two different types of growth rate $r$ , i.e., constant $r$ and time-varying $r$ are considered. The prediction errors are small compared with Diffusive Logistic model, which demonstrates the proposed model can well capture the trends of information diffusion for twitter.