محمدامين هوائي

Accurate prediction of building carbon emissions during early design remains a critical yet unresolved challenge for low-carbon construction policy. Traditional life cycle assessment (LCA) methods are often impractical at the permitting stage due to their reliance on detailed material inventories an‎d complex modeling workflows. To address this gap, this study proposes a novel transfer learning (TL) framework for predicting construction-stage carbon emissions using only minimal early-stage design inputs. Two TL scenarios are explored: (1) using detailed design data to predict material production carbon emissions from early-stage inputs, an‎d (2) predicting construction-phase emissions of one lifecycle stage based on material production stage. The framework combines stan‎dardized LCA-based carbon modeling with machine learning methods an‎d is validated using a dataset of 1,158 real-world construction projects in a major metropolitan city. The proposed TL models improve R² by +39.6% in Scenario 1 an‎d +36.8% in Scenario 2 compared to baseline models under limited-input conditions. To the best of the authorsʹ knowledge, this study represents the first application of transfer learning to carbon footprint prediction in the building sector, particularly within early-stage permitting workflows. The proposed approach enables data-efficient, scalable, an‎d lifecycle-aligned carbon estimation, offering actionable value to municipal permitting authorities, urban planners, an‎d design professionals seeking to integrate carbon intelligence into early decision-making.

پيش‌بيني ردپاي كربن , يادگيري انتقال , مجوزهاي ساختماني , كمي‌سازي عدم قطعيت , بهينه‌سازي طراحي پايدار

Carbon footprint prediction , transfer learning , construction permits , uncertainty quantification , sustainable design optimization

Mohammad Amin Havaei

Dr.Hassan Malekitabar