Meteorological Forecasting in Oman: A Comparative Analysis of Artificial Neural Networks, Support Vector Machines, Random Forest and Extreme Gradient Boosting Using Long-Term Historical Data (1984-2024)

Abstract This study presents a comprehensive comparative analysis of machine learning algorithms for maximum temperature forecasting in Oman using four decades of historical meteorological data from 1984 to 2024. Given the critical importance of accurate temperature prediction for solar energy output optimization in solar farms, we evaluated the performance of four prominent machine learning models: Random Forest, Extreme Gradient Boosting, Artificial Neural Networks, and Support Vector Machines. The models were trained on dataset incorporating geographical, temporal, and meteorological variables including longitude, latitude, elevation, precipitation and minimum temperature. The Random Forest model demonstrated superior performance with an RMSE of 2.09°C, MAE of 1.53°C, and R 2 of 0.92, followed closely by Extreme Gradient Boosting RMSE of 2.19°C, MAE of 1.64°C, R 2 of 0.91. The Artificial Neural Network achieved moderate performance with RMSE of 2.67°C, MAE of 1.94°C, and R 2 of 0.87, while the Support Vector Machine showed RMSE of 2.84°C, MAE of 2.01°C, and R 2 of 0.85. Computational efficiency analysis revealed Extreme Gradient Boosting as the fastest training model, while Random Forest provided the optimal balance between accuracy and computational efficiency. These findings provide crucial insights for solar energy planning and meteorological forecasting applications in arid and semi-arid regions, with implications for renewable energy infrastructure development in the Middle East.