The growing demand for energy in residential buildings poses significant challenges to achieving sustainability in urban development. Energy consumption, particularly for Heating Load (HL) and Cooling Load (CL), accounts for a substantial portion of residential building energy use. Therefore, predicting energy usage during the early design phases is crucial to making informed decisions that reduce long-term energy costs and environmental impacts. This study proposes a novel approach that leverages a Bayesian-based optimization technique to enhance the performance of the eXtreme Gradient Boosting (XGBoost) algorithm in predicting HL and CL. By optimizing the hyperparameters of the XGBoost model using Bayesian optimization, we improve the model’s ability to capture complex relationships between building design features—such as wall insulation, window size, geographical location, and climate conditions—and their associated energy consumption patterns. The results demonstrate that this hybrid approach provides highly accurate predictions, outperforming traditional energy estimation methods and offering a reliable tool for architects, engineers, and designers to optimize energy efficiency in residential building designs. The proposed method not only assists in better energy management but also supports the goal of sustainable building practices by providing early-stage insights into energy performance.

Keywords: Residential buildings, Energy usage prediction, Heating Load (HL), Cooling Load (CL), eXtreme Gradient Boosting (XGBoost), Bayesian optimization, Energy efficiency, Building design characteristics, Sustainable development, Machine learning, Prediction accuracy, Early design stage analysis.