A Comprehensive Railroad-Highway Grade Crossing Consolidation Model: A Machine Learning Approach

Area of Science:

• Transportation Engineering
• Data Science
• Public Safety

Background:

• Highway-rail grade crossings present significant safety risks, causing numerous vehicle-train incidents, injuries, and fatalities annually in the U.S.
• Current safety measures like grade separation and active alarms are employed, but crossing closures are the most effective safety improvement strategy.
• Incentive programs and consolidation models are crucial for facilitating the difficult process of closing highway-rail grade crossings.

Purpose of the Study:

• To develop a data-driven consolidation model for determining the suitability of highway-rail grade crossing closures.
• To utilize an eXtreme Gradient Boosting (XGBoost) algorithm for accurate prediction of crossing closure suitability.
• To identify key engineering variables influencing crossing closure decisions and provide insights into model behavior.

Main Methods:

• Application of an eXtreme Gradient Boosting (XGBoost) algorithm to a dataset of U.S. highway-rail grade crossings.
• Development of a data-driven consolidation model incorporating various engineering variables.
• Creation of a simplified model through sensitivity analysis, considering aggregate gain and correlation thresholds.

Main Results:

• The proposed XGBoost consolidation model achieved a high overall accuracy of 0.991 in predicting crossing closure suitability.
• The model identified the relative importance of input variables, offering transparency and understanding of its decision-making process.
• A simplified 14-variable model demonstrated similar performance to the full model, with 75% aggregate gain and a 0.9 correlation threshold.

Conclusions:

• The developed data-driven model effectively identifies highway-rail grade crossings suitable for closure, significantly enhancing safety.
• Approximately 62% of current highway-rail grade crossings are identified as candidates for closure based on the simplified model.
• The study provides a practical tool for transportation authorities to prioritize and implement effective safety improvement programs through strategic crossing closures.