AI-Driven Predictive Modeling for COVID-19 Case Trends: An Ensemble Learning Approach

Abstract

Pandemics and epidemics present major challenges to global health and economies, as underscored by the COVID-19 pandemic, which has highlighted the critical need for advanced predictive analytics to support informed decision-making and efficient resource allocation. In this context, Artificial Intelligence (AI) and Machine Learning (ML) have emerged as powerful tools for forecasting pandemic trends. This study investigates various ML regression techniques to predict COVID-19 case trends using real-world data from the Center for Systems Science and Engineering (CSSE) at Johns Hopkins University (JHU), covering the period from January 22, 2020, to September 3, 2023. The regression models evaluated include Linear Regression, Support Vector Regression (SVR), Random Forest, Gradient Boosting, AdaBoost, and a Stacking Model, with performance assessed using Mean Absolute Error (MAE), Mean Squared Error (MSE), and R² Score. Results reveal that the Stacking Model achieved the highest accuracy, with the lowest MAE (49,182,280), lowest MSE (3.32e+15), and best R² Score (-2.669640), outperforming all other models. Random Forest and Gradient Boosting also performed well with R² Scores of -4.857282 and -4.863941, respectively, while SVR proved unsuitable with an R² Score of -265.084156. These findings underscore the superior performance of ensemble learning techniques, particularly Stacking, Random Forest, and Gradient Boosting, in predicting COVID-19 trends and emphasize the importance of selecting appropriate regression models to enhance the reliability of epidemiological forecasting.