Predictive Modeling of Energy Requirements in the Design of Buildings: A Comparative Analysis of Machine Learning Algorithms

The demand for sustainable and energy-efficient buildings has grown significantly, making accurate prediction of energy loads a crucial aspect of modern building design. To address this need, machine learning algorithms have emerged as promising tools for precisely forecasting energy loads, thus helping to optimize energy consumption and reduce environmental impact. In this study, an in-depth comparative analysis of diverse Machine Learning Algorithms (MLAs) is presented. These MLAs include Linear Regression, Lasso Regression, Gradient Boost, and Decision Tree Regressors, and they are specifically tailored for predicting energy loads in buildings. To evaluate the predictive accuracy of each model, essential evaluation metrics such as Mean Absolute Error (MAE), Root Mean Squared Error (RMSE), and Coefficient of Determination (R-2) are employed. The study's findings highlight the exceptional performance of Gradient Boosting Regressor (GBR) among the selected MLAs, showcasing its superiority in accurately predicting energy loads. Additionally, the research underscores the significant impact of a building's Relative Compactness on its energy loads, emphasizing the importance of considering this factor in energy-efficient building design.