LEVERAGING AI-BASED PREDICTIVE SYSTEMS FOR REAL-TIME EPIDEMIC SURVEILLANCE AND CONTROL

Abstract

Real-time epidemic surveillance is essential for rapid detection, early intervention, and effective outbreak control, yet traditional surveillance systems struggle to capture rapidly evolving transmission dynamics. This study proposes an AI-based predictive framework designed to enhance epidemic monitoring by integrating multimodal data streams, including regional case trajectories, mobility patterns, symptom-reporting trends, and spatiotemporal risk indicators. The system employs a hybrid deep-learning architecture incorporating temporal modeling, nonlinear feature extraction, and attention-driven interpretability to forecast outbreak trajectories and detect potential hotspots. Experimental results demonstrate that the model consistently outperforms conventional forecasting approaches in capturing early transmission acceleration, identifying emerging high-risk regions, and generating stable probability-based outbreak predictions. Visualization of learned spatiotemporal patterns highlights the system’s ability to uncover hidden epidemic structures that traditional methods often fail to detect. Epidemiological expert review further confirmed that the model prioritizes meaningful predictors aligned with known transmission pathways and behavioral changes. While limitations exist due to data incompleteness, noise, and regional reporting variability, the findings provide strong evidence that AI-driven predictive systems can significantly improve real-time epidemic intelligence, enabling faster response, better allocation of public-health resources, and more accurate early-warning capabilities. This study contributes a scalable and interpretable AI framework for modern epidemic surveillance, laying the foundation for precision-driven outbreak prediction and proactive health-system preparedness.