AI-driven Social Network Analysis for Epidemic Diagnosis and Asymptomatic Infector Tracking

Abstract

The widespread dissemination of infectious diseases, such as COVID-19, has underscored the importance of early epidemic detection and tracking asymptomatic infectors within interconnected populations (e.g., schools). Classical compartmental models offer valuable insights but sometimes miss the implicit social network structure that promotes transmission. This paper proposes an artificial intelligence-enhanced social network analysis (SNA) framework that combines epidemiological simulations and machine learning (ML) methodologies for epidemic detection and intervention evaluation. We start by applying the publicly available SocioPatterns high school dataset to create aggregated, multilayer contact networks. We further model different Susceptible-Exposed-Infectious-Recovered (SEIR)-based intervention scenarios (the baseline, full isolation, partial isolation, and a combination of both) to assess epidemic dynamics and class-specific attack rates. To predict potential infectors, we utilise gradient boosting and Node2Vec embeddings with logistic regression. SHapley Additive exPlanations (SHAP) values also aid in interpreting feature importance. Finally, we propose a fusion method combining handcrafted and embedding-based features, showing better predictive performance. Our results demonstrate that our framework can generate realistic epidemic patterns and facilitate the early identification of asymptomatic carriers, thereby providing a guideline for dormitory-based educational institutions to respond to the pandemic. © 2026 IEEE.