Enhancing COVID-19 Forecasting Precision through the Integration of Compartmental Models, Machine Learning and Variants

Daniele Baccega, Paolo Castagno, Antonio Fernández Anta, Matteo Sereno

Keywords: Artificial Intelligence, Epidemics, Compartment models, Variants, Forecasting, COVID-19

Abstract: Predicting epidemic evolution is essential for making informed decisions and implementing effective countermeasures. Computational models provide valuable insights into disease progression, enabling early detection, timely intervention, and effective prevention strategies. These models help allocate resources and protect public health by anticipating the course of an outbreak and allowing for proactive measures. We propose Sybil, a framework that merges machine learning with variant-aware compartmental models, combining data-driven and analytical methods. We tested Sybil's predictive capabilities using COVID-19 data from Italy, Austria, and U.S., including records of new and recovered cases, fatalities, and the presence of different variants over time. Our evaluation focused on Sybil's forecasting accuracy during periods of significant trend changes. The results indicate that Sybil surpasses traditional data-driven approaches, accurately predicting trend shifts and the extent of these changes.

Submission Number: 12