A simulation-based approach to predicting influence in social media communities: A case of U.S. border security

Predicting influence in social media (SM) communities has a strong implication for cybersecurity and public policy setting. However, the rapidly growing volume and large variety of SM have made the prediction difficult. Unfortunately, research that combines the power of simulation, SM networks, and SM community features to predict influence is not widely available. In this research, we developed and validated a simulation-based approach to predicting influence in SM communities. The approach uses a power-law distribution to simulate user interaction and leverages statistical distributions to model SM posting and to predict influence of opinion leaders. We applied the approach to analyzing 1,323,940 messages posted by 380,498 users on Twitter about the U.S. border security and immigration issues. Three models for predicting behavioral responses were developed based on exponential distribution, Weibull distribution, and gamma distribution. Evaluation results show that the simulation-based approach accurately modeled real-world SM community behavior. The gamma model achieved the best prediction performance; the Weibull model ranked second; and the exponential model had a significantly lower performance. The research should contribute to developing a simulation-based approach to characterizing SM community behavior, implementing new models for SM behavior prediction, providing new empirical findings for understanding U.S. border security SM community behavior, and offering insights to SM-based cybersecurity.