Predicting Sustainability Performance in Construction Projects Using

The construction sector plays a major role in global environmental degradation, contributing significantly to carbon emissions, energy consumption, and waste generation. Despite the urgency to address these challenges, limited studies have explored the integrated prediction of sustainability performance using real-world project data, particularly in the context of Saudi Arabia. This study aims to bridge this gap by applying supervised machine learning techniques to predict carbon emissions and classify projects based on their emission levels. A structured survey was conducted, and 150 validated responses from key project stakeholders across Saudi Arabia were collected, covering a wide range of project and sustainability parameters. Three machine learning models, Support Vector Machine (SVM), Random Forest (RF), and Extreme Gradient Boosting (XGB) were trained and evaluated. Using 10-fold cross-validation on the training set, XGB achieved the highest mean. In classification, both XGB and SVM achieved the highest accuracy of 76%, while RF followed with 73%. SHAP analysis revealed that waste generation, energy consumption, and project duration were the most influential predictors of carbon emissions. The findings offer a practical machine learning framework for early sustainability assessment and policy planning aligned with Saudi Vision 2030.