Modelling and Analyzing Social Interactions in COVID-19 with Dynamic Graphs

Social networks have become significant communication platforms for tracking user sentiments, interests, and activities, particularly during the COVID-19 pandemic when individuals have been confined to their homes. While static social contact networks during COVID-19 have been widely studied, the dynamic characteristics of social interactions have not been thoroughly investigated. In this paper, we model social interactions using dynamic graph methods and apply the proposed approach to analyze the social interactions within a Weibo dataset. Moreover, characteristics of dynamic graphs including node degree distributions are studied. Our analysis illustrates that social interaction frequencies are increasing and social interaction communities are forming as the COVID-19 pandemic continues to evolve. Furthermore, we show that the trend of social interaction frequencies is consistent with the trend of infected person numbers in the real world as compared by KL divergence. These results suggest that interventions targeting users with higher social interaction frequencies and users belonging to the same communities might aid in the development of effective epidemic prevention policies.