Multi-source information diffusion in online social networks

Individual spreading behavior in online social networks is closely related to user activity, tie strength, and other user and network features. The results concentrate on personal spreading decisions; however, whether these features promote the global information diffusion and increase the macroscopic density of infected agents, remains unclear to us. In this paper, we propose a multi-source diffusion model in which agents may create new messages and spread other agents’ messages. Agents receive many messages, and each time they select a certain message preferentially to spread in consideration of different features. Simulation results show the density of infected agents for different messages follows a power-law distribution in both scale-free and small-world networks. Selecting the largest author degree, author activity and tie strength preferentially can advance the overall diffusion process. Weak tie bias is the least effective feature for multiple information diffusion, but it helps to diffuse a single message. Unexpectedly, the bias of interest similarity does not have an apparent effect. Integrated with the influence on individual diffusion behavior, strong tie bias is a significant feature both for local and global diffusion.