How does information diffuse in large recommendation social networks?

This article investigates how information diffuses in Douban, an online social network. First, we analyze properties of the user relationships in Douban, observing its degree distribution, network reciprocity, and degree of separation. Second, we propose a method that infers how the information diffuses through this network. Subsequently, using this method we rebuild possible information diffusion graphs, and make statistical observations of the disconnected properties, size distributions, and diffusion patterns in Douban. Based on our empirical analysis, we found that in most cases, information diffuses with multiple origins in Douban. We attribute this to the two different kinds of influence that lead to the diffusion: internal and external influence. Finally, based on the observations, we propose a novel SID model to formulate the diffusions in online social network environments. In the model, there are n iterations. On each iteration, each non-infected node can be infected, either externally with probability P_ex, or internally with probability P_in from each of the nodes it follows that were already infected. Our simulation results reveal that the SID model can flexibly portray the diffusion processes in Douban through adjusting these two probabilities.