Optimized machine learning for building energy prediction: Feature selection and hyperparameter tuning using the Al-Biruni Earth Radius (BER) search optimization algorithm

Building energy consumption constitutes a significant portion of global energy demand, highlighting the need for accurate predictive models to enhance energy efficiency, reduce operational costs, and support sustainable urban development. In response to this need, this study proposes an optimized machine learning framework for predicting Site Energy Usage Intensity (Site EUI), utilizing a dataset comprising building characteristics, weather conditions, and historical energy usage data. The primary contribution of this work lies in integrating the Binary Al-Biruni Earth Radius optimizer (bBER) with several metaheuristic algorithms—namely Grey Wolf Optimizer (GWO), Particle Swarm Optimization (PSO), and Whale Optimization Algorithm (WAO)—to enhance both predictive accuracy and computational efficiency. Experimental results demonstrate that the baseline Random Forest model, without feature selection, yielded an RMSE of 0.1096 and a coefficient of determination ( R 2 ) of 0.8258. Incorporating bBER-based feature selection reduced the RMSE to 0.0183 and improved R 2 to 0.9024. Further optimization using the BER-based Random Forest model achieved an RMSE of 0.000983 and R 2 of 0.9843. These findings confirm that metaheuristic-driven feature selection and optimization significantly improve prediction accuracy and reduce model complexity, making the proposed framework a valuable tool for intelligent building energy management and the advancement of energy-efficient smart grids.