Unlocking Sustainable Future: Machine Learning for Precise Energy Efficiency Prediction in Buildings

To lessen the negative effects on the environment and ensure sustainable development, the building sector urgently needs to switch to energy-efficient designs. The size, purpose, and structure of buildings vary widely. It is difficult to create a single model that generalizes successfully across multiple building types. The consumption of energy in buildings can be significantly impacted by external variables such as weather and seasonal changes. It is not easy to incorporate these aspects into predictive models appropriately. This research aims to maximize energy usage through the techniques of machine learning methods to assess the heating and cooling loads in buildings. The UCI Machine Learning Repository’s Energy Efficiency dataset is the basis for developing predictive models that can calculate the heating and cooling loads depending on various building attributes. Before model training, considerable feature engineering and data preprocessing are carried out, including addressing missing values, scaling numerical features, and encoding categorical variables. Insights into the connections between attributes and goal variables are discovered through exploratory data analysis. Five well-known machine learning regression algorithms based on Decision Tree, Support Vector Linear, and Gradient boosting are employed in the study. Each model is trained using the training dataset, and cross-validation methods to adjust the hyperparameters. The model performance is assessed using the metrics mean squared error (MSE), coefficient of determination (R-squared), and mean absolute error. The study can suggest the most reliable and precise forecast model that can estimate the heating and cooling demands. This work advances sustainable construction practices by demonstrating the effectiveness of machine learning in energy efficiency prediction.