Exploring the Use of Machine Learning in Enhancing BiddingDecisions for Construction Projects

Abstract

This study addresses the critical need to enhance competitive bidding strategies in construction by revisiting Friedman's 1956 bidding model and incorporating Ioannou's revised equations to improve predictive accuracy. The research fills a significant gap in integrating machine learning techniques into bidding theory, which offers a data-driven approach to optimize bid decision-making. Using synthetic data, logistic regression served as a baseline model, while Random Forest classifiers outperformed with 98% accuracy by addressing class imbalance and effectively capturing the non-linear relationships among key variables, such as reserve price and bid-to-cost ratio. The findings revealed that machine learning models could simplify complex bidding theories and provide contractors with actionable insights, supporting bid or no-bid decisions. However, reliance on synthetic data limits the generalizability of these results. Future work should focus on validating the proposed models using real-world bidding datasets and exploring advanced techniques, such as ensemble methods, to enhance predictive performance. This study underscores the potential of machine learning to transform traditional bidding practices, which offers both theoretical advancements and practical implications for construction management.