Residential Building Energy Performance Analysis Using Machine Learning Algorithms

The HVAC system can achieve low energy consumption as ML-based models accurately estimate the building's energy use and load demands. Therefore, in this work, an extreme gradient boosting (XGBoost) ensemble model is proposed for predicting energy usage based on heating and cooling Loads (HL and CL). Furthermore, RF, LR, KNN, and SVR are also designed for comparison analysis. The results show that XGBoost outperforms all the applied algorithms, achieving the lowest values of RMSE (0.407 and 0.858) and MSE (0.166 and 0.737) in both cases. Furthermore, its performance is also compared with the models presented in the literature. Finally, it can be concluded that the proposed XGBoost is superior, robust, and efficient for predicting HL and CL, respectively.