Comparing AI-Based and Traditional Epidemiological Models for Epidemic Forecasting in the USA. A Comparative Study of AI Models Versus Conventional Statistical Public Health Methods.

Abstract

Artificial intelligence has emerged as a transformative force in epidemic forecasting, offering new opportunities to complement and extend traditional epidemiological modeling approaches. This study presents a comparative analysis of AI-based models and conventional public health forecasting methods for epidemic prediction in the United States. The research examines the performance and practical implications of mechanistic and statistical techniques alongside modern machine learning frameworks within a unified analytical setting. The methodological framework incorporates traditional epidemiological models, including SEIR and ARIMA approaches, together with Random Forest and Long Short-Term Memory architectures representing contemporary artificial intelligence methods. A standardized preprocessing pipeline, common evaluation criteria, and consistent experimental procedures are employed to ensure fair comparisons across forecasting paradigms. Beyond predictive accuracy, the study considers interpretability, robustness, computational efficiency, and operational applicability as critical dimensions of effective epidemic forecasting. The findings indicate that AI-driven models exhibit stronger capabilities in capturing nonlinear transmission patterns and adapting to rapidly evolving epidemic conditions. Deep learning methods, particularly recurrent architectures, demonstrate enhanced performance during complex outbreak waves characterized by changing behavioral and intervention dynamics. Traditional epidemiological approaches, however, retain important advantages in interpretability, transparency, and policy communication, providing mechanistic insights that remain essential for public health decision-making. The results further suggest that methodological diversity contributes to greater forecasting resilience and supports more reliable epidemic preparedness strategies. The study concludes that artificial intelligence should not be viewed as a replacement for conventional epidemiological modeling but rather as a complementary set of tools capable of strengthening existing public health infrastructures. Future forecasting systems are likely to benefit from hybrid and collaborative frameworks that integrate mechanistic understanding with data-driven learning capabilities. Such an approach offers a promising pathway toward more accurate, interpretable, and operationally effective epidemic forecasting in the United States.