Machine learning models as a budget validation method for U.S. mass-transit projects

Capital budgets for U.S. mass transit projects have consistently been underestimated. The resulting budget overruns cause funding challenges for urban transit authorities and the Federal Transit Administration. Developing accurate budget estimates is a significant challenge for project managers, particularly given the influence of cognitive biases such as optimism bias.,This study developed a data-driven approach that applies supervised and unsupervised machine learning techniques to improve the reliability of cost estimates. A dataset of 108 federally supported rail-transit projects was analyzed using neural networks and k-means clustering in standard analytical software. The results demonstrated that the approach produces statistically significant improvements in predictive accuracy and provides a transparent, reproducible method for early cost assessment. The findings offer a practical solution for early-stage budget validation that reduces bias in mass transit project planning.