Proposal for machine learning-based decision support system for contractor selection in public contracts

The construction sector is a sector where complex projects with high risks are carried out, working with different fields of expertise that are closely connected with other sectors through both private and public. The public is in a kind of an employer position in the construction sector by procuring some of the works it needs from the private sector through tenders. The public procurement process is represented by a set of rules and procedures that determine how procurement activities are conducted. The Public Procurement Law documents developed to be used in public construction works tenders are used in achieving the determined cost, duration, and quality targets of construction projects. In public tenders, a selection process based mainly on the most economically suitable bid price is followed. This method, which is applied, causes some difficulties in completing the construction projects in the desired time, budget, and quality. Although it is desired to try to prevent this situation by taking into consideration the non-price elements, sufficient importance is not given to the qualification criteria of the bidders in the bid evaluation process. In this study, the data of the completed public works were examined, and decision-support models were created by using machine learning algorithms to estimate the cost to be completed with the desired performance or achieve the desired performance. The performance results of the models were compared with the Accuracy, Sensitivity, Precision, F1 Score metrics. As a result, the employer will be able to predict the success class of the contractor with the penalty rate to be determined by considering different factors such as complexity, risk, work size, etc. of the project. Keywords: Tender, contractor selection, machine learning algorithms