Machine Learning-Enhanced Building Information Modeling (BIM) for Sustainable Urban Planning
Rehab Salaheldin Ghoneim, Racha Taiyara, Zain Nader Maghaireh
Abstract
The main concern in the modern world is global energy consumption and environmental impacts; thus, sustainable architecture is a significant concern. For this, advanced predictive models become inescapable for better optimization of design outcomes. This research investigates using CNNs to further the sustainability of buildings through precise predictions of critical metrics such as energy efficiency, carbon footprint, and occupant satisfaction. The best CNN model showed high accuracy: 91.2% in energy efficiency classification, with a low MSE of 0.014 for energy predictions and 35.8 tons CO<sub>2</sub> for carbon footprint estimates. It also reached an accuracy of 88.5% in predicting occupants' satisfaction with a 0.023 MSE. Coupling CNN predictions with BIM workflows allows architects to work on material selection, building footprint, and urban density in the earliest design stage by promoting proactive sustainability measures. The study, however, is very effective at realizing limitations in both the diversity of the dataset used and the static nature of the design focus and points toward future enhancements in expanded datasets, dynamic data integration, and improved computational efficiency. This research identified CNN as an essential tool concerning sustainability design optimization that is appropriate for aligning architectural practice with energy efficiency and environmental goals.