<![CDATA[The rising energy demands of buildings necessitate systems for efficient energy usage. Consequently, accurate prediction methods are required to assist in energy planning, control, and optimization. This study aims to predict energy consumption using the XGBoost algorithm, with parameter optimization performed via Grid Search. Additionally, Random Forest is employed as a benchmark and comparative method, and feature importance analysis is conducted to identify the variables that most significantly influence energy consumption. The dataset used is a synthetic dataset obtained from Kaggle. The dataset simulates environmental conditions and energy use in buildings. The results indicate that the best-performing model is the parameter-optimized XGBoost, which outperformed both Random Forest and the default XGBoost model. Specifically, the optimized XGBoost model achieved an MAE of 4.0029 ± 0.0127, an RMSE of 4.9352 ± 0.0104, and an R² of 0.6026 ± 0.0017. Feature importance analysis identified temperature as the most important feature for predicting energy consumption in the evaluated model. However, because the study is using synthetic data rather than real data from actual building measurements, the findings may not fully represent real-world buildings and should not be directly generalized without further validation. Further evaluation using real-world building energy consumption data is necessary to assess the practical applicability of the approach.]]>
Copyrights © 2026
