Abstract

This study proposes a novel Deep Epidemiological Resource Intelligence Architecture (DERIA) for intelligent data-driven hospital-system management under dynamic epidemic environments. The proposed framework integrates nonlinear epidemiological modeling, transformer-based temporal healthcare learning, graph neural interaction analysis, Bayesian uncertainty forecasting, and reinforcement-learning-driven resource optimization into a unified computational ecosystem. Unlike conventional healthcare-management systems, the proposed methodology establishes a bidirectional interaction mechanism between disease propagation and hospital operational dynamics. The framework was evaluated using large-scale hospital and epidemiological datasets containing approximately 1.8 million healthcare observations and 32 operational features. Experimental results demonstrated superior predictive and operational performance with 98.1% prediction accuracy, 97.4% precision, 96.8% recall, and significant reductions in healthcare congestion and mortality risk. Furthermore, the proposed framework achieved 97.8% ICU resource-utilization efficiency and 84.7% emergency-congestion reduction. The findings confirm that the proposed DERIA framework provides a scalable, adaptive, and intelligent solution for epidemic-aware hospital management and healthcare-resource optimization.