Independent variables combination selection using best subset selection method in a multiple linear regression baseline energy model for educational building’s energy consumption prediction

A baseline energy model (BEM) establishes a relationship between energy consumption and its governing independent variables, serving as a foundation for predicting energy usage. Typically, baseline energy models often rely on multiple linear regression due to its simplicity and effectiveness in estimating energy consumption based on selected variables. However, traditional baseline models may suffer from reduced performance when too many independent variables are included, as not all variables have a strong impact on energy consumption. This can lead to overfitting and decreased predictive accuracy. To address this issue, this project introduces an enhanced baseline energy model that integrates the best subset selection method. This approach identifies the most impactful independent variables, ensuring a more accurate and efficient model for energy consumption prediction. The enhanced model demonstrates superior performance, with a mean squared error (MSE) of 255.00 kWh compared to 255.34 kWh to the traditional model. This improvement highlights the model’s ability to choose relevant variables, delivering better prediction accuracy. The model offers significant advantages, including improved energy-saving planning and operational optimization. With strong commercialization potential, it can be applied to buildings with similar characteristics, fostering sustainable energy management and contributing to socio-economic and environmental benefits.