Trajectory Linkage and Spreader Centrality for Social Epidemic Networks

Constructing social networks for real epidemic cases is very challenging. Many mathematical models have been proposed to model such networks using simulation models, such as susceptible-infected (SI), susceptible-infected-recovered (SIR), and susceptible-exposed-infectious-removed (SEIR). Nonetheless, social network analyses can fail to capture real conditions, as such models are constructed based on many assumptions. Furthermore, unlike standard online social networks (OSNs), a social epidemic network requires different treatment from both model construction and social network analysis perspectives, especially for the detection of superspreaders. To address these issues, we propose a trajectory linkage method to automatically discover social networks from historical patient-trajectory data, wherein relations among patients are determined by spatial proximity and time-windows. Moreover, we introduce a novel spreader centrality measure that is devised to identify superspreaders in a social epidemic network. Extensive experiments were performed using real epidemic data. The results revealed that trajectory linkage can obtain a denser social network model than is possible by only incorporating patient data (the “who is infected by whom” relationship). By performing a social network analysis, the trajectory linkage model can express the real conditions of the patient relationship. Furthermore, our spreader centrality can capture the real superspreaders more effectively than can the existing centrality measure in social epidemic networks.