User characteristics based information diffusion model for analysis of hot social events
In this paper, we studied the statistical features of user posting behavior in microblog. The number of statuses exhibits a double power-law distribution and there is strong positive correlation between the number of forwarding and comments. According to the empirical analysis of user behavior in both group level and individual level respectively, we find the interval distribution of user posting behavior follows a power-law distribution and there is positive correlation between users' activity and the power-law exponent. There is obvious periodicity of the user posting time series to some extent. Further studies show that user's interests are influenced by the forwarding and comment behaviors. Considering these effects, this paper analyzed an improved user posting behavior model co-driven by user's interests and interactions. We also proposed another model where user's interests change with Logistic function. The simulation results verify these two models and provide reference to the studies of human dynamics in online social networks.