<![CDATA[The growing demand for energy-efficient buildings requires accurate predictive models to support sustainable development. However, conventional prediction approaches often struggle to capture the complex nonlinear relationships between building design characteristics and energy consumption, limiting prediction accuracy for practical building energy management. Therefore, developing more reliable predictive models has become an important challenge in supporting data-driven energy management. This study implemented and evaluated machine learning models, specifically ensemble learning methods, to predict building energy efficiency based on building design parameters. The study aimed to compare the predictive performance of conventional regression and ensemble learning models for estimating heating and cooling loads while identifying the most influential building design parameters affecting energy efficiency. The proposed methodology consisted of data preprocessing, model development using Linear Regression, Random Forest, and Gradient Boosting, followed by performance evaluation using Mean Absolute Error (MAE), Mean Squared Error (MSE), and the coefficient of determination (R²). The results indicate that the ensemble learning models substantially outperform Linear Regression in predicting both heating and cooling loads. Random Forest achieved the best performance for heating load prediction, while Gradient Boosting performed best for cooling load prediction. Feature importance analysis showed that geometric parameters, including relative compactness, overall height, and surface area, had the greatest influence on energy efficiency. These findings demonstrate that ensemble learning not only improves prediction accuracy but also enhances model interpretability through feature importance analysis, providing valuable support for data-driven decision-making in smart and sustainable building systems.]]>
Copyrights © 2026
