Leveraging machine learning for data-driven building energy rate prediction

• Published in: Results in engineering Vol. 26; p. 104931
• Main Authors: Eslamirad, Nasim, Golamnia, Mehdi, Sajadi, Payam, Pilla, Francesco
• Format: Journal Article
• Language: English
• Published: Elsevier B.V 01.06.2025; Elsevier
• Subjects: Building energy rate (BER); Data-driven approach; Machine learning classification models; Urban building energy modelling (UBEM); Urban energy analysis; Urban energy analysis; Urban building energy modelling (UBEM); Machine learning classification models; Data-driven approach; Building energy rate (BER)
• ISSN: 2590-1230; 2590-1230
• DOI: 10.1016/j.rineng.2025.104931

•Proposes a novel data-driven methodology for predicting Building Energy Ratings (BER) in urban areas, with Dublin, Ireland, as a pilot case.•Integrates geospatial, building-specific, and neighborhood-scale environmental data for BER modeling.•Utilizes advanced Machine Learning algorithms (RF, DT, KNN, SVM) to achieve high predictive accuracy.•Aligns BER prediction framework with European energy performance standards, ensuring scalability across urban contexts.•Offers a practical tool for policymakers to enhance urban energy planning and reduce carbon emissions, setting a new benchmark for predictive accuracy in Urban Building Energy Modeling (UBEM). This paper presents a novel, data-driven approach for predicting Building Energy Ratings (BER) in urban environments, using advanced Machine Learning (ML) algorithms. Focusing on Dublin, we integrate diverse geospatial datasets with building-specific and neighbourhood-scale features to classify BER. Our approach leverages cutting-edge ML techniques, including Decision Trees (DT), Random Forest (RF), K-Nearest Neighbours (KNN), and Support Vector Machines (SVM), to develop highly accurate predictive models. The performance of these models was rigorously evaluated using comprehensive statistical metrics, such as Receiver Operating Characteristic (ROC), Area Under the Curve (AUC), precision, recall, and overall accuracy (OA). The results demonstrate that the RF model outperformed other algorithms, achieving the highest classification accuracy, with AUC values ranging from 0.77 to 0.83. Notably, the model exhibited superior performance in classifying higher energy-consuming buildings (Class G), while the SVM showed limited discriminative power with AUC values between 0.37 and 0.49. These findings underscore the effectiveness of ML in Urban Building Energy Modelling (UBEM), particularly in forecasting energy consumption patterns and identifying high-energy-use buildings. This study makes significant contributions by advancing the application of ML to urban energy efficiency planning. By aligning the Building Energy Rating (BER) prediction framework with European energy performance standards (ISO/CEN), it ensures adaptability and relevance across diverse urban contexts. The approach addresses key limitations in UBEM while offering a robust tool for policymakers and urban planners to optimize energy consumption and reduce carbon emissions. Integrating spatial and contextual factors with BER establishes a new standard for predictive accuracy in urban energy research.