Background. The recent global pandemic created unique difficulties for healthcare systems all around the world that required accurate predictions to make decisions about resources management and other factors. For Saudi Arabia, an essential component for managing a crisis like this includes analyzing the dynamics of changes within case numbers. Aims. This paper seeks to improve healthcare decision-making by conducting a comparative assessment of the prediction accuracy of various machine learning models used for forecasting COVID-19 cases in Saudi Arabia. Methodology. Five ML models will be considered in this study – Decision Tree Regressor, Random Forest Regressor, LSTM, CNN, and SARIMAX. The selected models will be trained using the series of time-related data concerning the number of patients confirmed with coronavirus in Saudi Arabia. Models’ performance will be assessed based on MAE, RMSE, and R2 metrics. Results. The results revealed Random Forest as the most accurate model with a set of metrics being (MAE: 0.214106, RMSE: 0.809336, R2: 0.999997). The second place went to Decision Tree with MAE equaling 0.466465, RMSE equaling 1.168835, and R2 equaling 0.999976. As for the third and fourth places, both deep learning models achieved comparable metrics and significantly outperformed ARIMA and SARIMAX. Conclusion. Machine learning models have proven their efficiency in such applications. The study identified a need for region-specific predictive models. It also pointed to a necessity to integrate data and create a model with greater accuracy and speed of processing.
Background. The recent global pandemic created unique difficulties for healthcare systems all around the world that required accurate predictions to make decisions about resources management and other factors. For Saudi Arabia, an essential component for managing a crisis like this includes analyzing the dynamics of changes within case numbers. Aims. This paper seeks to improve healthcare decision-making by conducting a comparative assessment of the prediction accuracy of various machine learning models used for forecasting COVID-19 cases in Saudi Arabia. Methodology. Five ML models will be considered in this study – Decision Tree Regressor, Random Forest Regressor, LSTM, CNN, and SARIMAX. The selected models will be trained using the series of time-related data concerning the number of patients confirmed with coronavirus in Saudi Arabia. Models’ performance will be assessed based on MAE, RMSE, and R2 metrics. Results. The results revealed Random Forest as the most accurate model with a set of metrics being (MAE: 0.214106, RMSE: 0.809336, R2: 0.999997). The second place went to Decision Tree with MAE equaling 0.466465, RMSE equaling 1.168835, and R2 equaling 0.999976. As for the third and fourth places, both deep learning models achieved comparable metrics and significantly outperformed ARIMA and SARIMAX. Conclusion. Machine learning models have proven their efficiency in such applications. The study identified a need for region-specific predictive models. It also pointed to a necessity to integrate data and create a model with greater accuracy and speed of processing.
