ENHANCED MACHINE LEARNING DECISION SUPPORT SYSTEMS FOR COVID-19 MANAGEMENT IN SAUDI ARABIA: A COMPARATIVE STUDY OF ADVANCED PREDICTIVE MODELS

Abstract

This study analyzes the usefulness of several machine learning (ML) models in predicting COVID-19 cases in Saudi Arabia, with the goal of improving healthcare decision-support systems. The comparative analysis includes five major models: Decision Tree Regressor, Random Forest Regressor, Seasonal Autoregressive Integrated Moving Average with Exogenous Inputs (SARIMAX), Long Short-Term Memory (LSTM), and Convolutional Neural Network (CNN). The Random Forest model had the highest predictive accuracy (MAE: 0.214106, RMSE: 0.809336, R²: 0.999997), followed by the Decision Tree Regressor. In contrast, classic statistical approaches such as ARIMA and SARIMAX performed significantly worse, suggesting their inadequacies in the setting of COVID-19. The study emphasizes the importance of specialized modeling approaches that take into account local epidemiological characteristics and encourages continued interactions between data scientists and healthcare practitioners. Furthermore, the findings call for greater data integration, model validation, and training for healthcare workers in order to maximize the use of sophisticated prediction models. Overall, this study provides significant information for healthcare officials, allowing them to make more educated decisions and allocate resources strategically when dealing with public health problems.