<![CDATA[This study addresses the necessity of evaluating solar radiation as a renewable energy source in tropical regions, specifically focusing on the challenges of estimation in Batam. The objective is to model daily solar radiation levels using satellite-derived weather data to overcome the lack of surface observation stations. Daily meteorological variables, including air temperature, relative humidity, rainfall, surface pressure, and wind speed, were sourced from the NASA POWER platform for the period January 1, 2020, to July 2, 2025. To ensure robust model generalization and prevent data leakage, the dataset was partitioned chronologically, utilizing data from 2020–2024 for training and the year 2025 for independent testing. Three computational models Linear Regression (LR), Random Forest (RF), and Decision Tree (DT) were applied to the processed data. The evaluation results indicate that the Random Forest model achieved the highest relative performance among the tested algorithms, recording a Mean Squared Error (MSE) of 19.61, a Mean Absolute Error (MAE) of 3.42, and a coefficient of determination R² of 0.20. In comparison, the Linear Regression model produced an R² of 0.19, while the Decision Tree showed significantly lower predictive accuracy. Despite being the most viable model, an R² of 0.20 reveals that the current predictors explain only 20% of the variance in solar radiation, highlighting the inherent complexity of tropical atmospheric dynamics. These findings suggest that while machine learning offers a promising framework for energy planning in Batam, further research incorporating additional explanatory features, such as cloud cover or aerosol indices, is required to improve model reliability.]]>
