This research aims to improve the accuracy of individual freight train run time predictions. A
regression model is proposed utilizing a broad selection of explanatory variables. The performance of the proposed regression model is compared against a baseline simple historical averaging technique. The proposed regression model offers substantial improvements in accuracy over the baseline technique: 36.79%, 28.74%, and 20.95% for low, medium, and high priority trains, respectively. The model justifies further exploration by the partner railroad to enable more accurate train schedules with subsequent improvements in railroad capacity, customer service, and asset utilization.
Authors: Kunal Bonsra and Joseph Harbolovic
Advisor: Dr. Basak Kalkanci and Prof. Eva M. Ponce Cueto