Machine Learning-Based Prediction of Energy Consumption in Smart Buildings for Sustainable Energy Management

Abstract

Sustainable operations together with higher efficiency depend on proper energy management in buildings with smart functionality. In this paper, the machine learning methods using multiple regression and advanced learning models are employed to predict the energy consumption (EC). The dataset consists of key input variables, including temperature (T), humidity (H), occupancy (O), building area (A), lighting power usage (L), HVAC energy consumption, and the day of the week (D). The proposed methodology involves data preprocessing, feature selection, and hyperparameter tuning to enhance model accuracy. Four machine learning models—multiple linear regression (MLR), random forest regressor (RFR), support vector regression (SVR), and artificial neural networks (ANN)—were evaluated based on performance metrics such as R², mean squared error (MSE), and root mean squared error (RMSE). Results indicate that the random forest regressor outperforms other models, achieving an R² of 0.84 and an accuracy of 89.27% on the test data, while ANN, despite excelling in training, demonstrated overfitting with reduced generalization ability. Sensitivity analysis highlights HVAC energy consumption and lighting power usage as the most influential parameters. An actual case study illustrates how the implemented model functions in modern-day energy management practice by demonstrating energy saving opportunities along with proposed optimization measures. These findings help sustainable energy practices through predictable choice making and enhanced energy efficiency monitoring of intelligent building operations.