Modeling Retweets’ Network Structure, Growth Process, and Frequency Time Trajectory: Their Correlations and Implications
Information diffusion via social media is often conceptualized as a network which can be modelled by at least three different ap-proaches: network structure (mainly using social network analysis), network growth process (power-law distribution), and frequency-time analysis (e.g. SIR model). This study aims to bridge these three approaches, derive their corresponding model parameters, find their correlations, and test the models against empirical real-life data. Using API provided by a microblog service provider in China, this study collected 100 complete sets of real-life networks of retweets about a variety of social problems in China. The net-work data were then analyzed by using social network analysis and power-law distribution fitting. Their frequency-time trends were modelled by a modified version of SIR (Susceptible, Infective, and Recovered) model which is customized for retweeting. Correla-tional analysis was conducted to find associations between the three sets of model parameters.