A machine-learning-based framework for contractor selection and order allocation in public construction projects considering sustainability, risk, and safety

Effective contractor selection is crucial for successful execution of construction projects. In contrast to the conventional lowest-bid approach prevalent in the public sector, this study focuses on developing a framework that minimizes time and cost overruns by considering diverse criteria for contractor selection. A variety of machine learning models, including multi-linear regression, random forest, Support Vector Machine, and Artificial Neural Network, have been employed, with multi-linear regression proving to be the most effective, achieving the lowest Mean Squared Error of 0.00003366. To determine the final order allocation, a multi-objective mathematical model was utilized to optimize conflicting criteria, such as time and cost overruns, sustainability, risk, and safety aspects related to shortlisted contractors. The findings highlight the significance of specific selection criteria, such as turnover, experience in similar projects, qualification of staff, technology utilization, client satisfaction, accident records, available bid capacity, and socioeconomic factors. This study emphasizes a three-phase decision-making framework for contractor selection and order allocation, particularly in public construction projects, with a focus on sustainability. By adopting this approach, government agencies can enhance infrastructure projects and minimize overruns through optimization and analytical tools, which aligns with the Gati-Shakti scheme of the Indian government. It is recommended that clients embrace a holistic approach to contractor selection, considering both technical and non-technical factors, to ensure successful project outcomes.