Modeling highway-rail grade crossing (HRGC) crash severity using statistical and machine learning methods.

Area of Science:

• Transportation Safety
• Traffic Engineering
• Accident Analysis

Background:

• Highway-rail grade crossing (HRGC) safety is a critical concern, with crash severity being a primary issue.
• Existing research often overlooks region-specific factors, necessitating localized analysis of HRGC crash determinants.

Purpose of the Study:

• To model HRGC crash severity using statistical and machine learning methods.
• To identify significant factors contributing to severe injury outcomes in HRGC crashes in Florida.

Main Methods:

• Utilized five years (2017-2021) of crash data from Florida's state-maintained HRGCs.
• Employed Ordinal Logistic Regression (OLR) and Random Forest (RF) algorithms for statistical and machine learning modeling.
• Analyzed variables including roadway design, driver behavior, environmental conditions, and crash characteristics.

Main Results:

• OLR identified ten significant variables, with factors like high speed limits, significant vehicle damage, and specific driver actions increasing crash severity.
• RF model highlighted estimated vehicle damage, speed limit, driver action, and crash type as crucial factors.
• Both models indicated that factors such as high speed limits, vehicle damage, and driver actions are strongly associated with increased injury severity.

Conclusions:

• Identified critical factors influencing HRGC crash severity, including vehicle damage, speed limits, and driver actions.
• Results provide valuable insights for developing targeted countermeasures to reduce fatalities and injuries at HRGCs.
• The study emphasizes the importance of considering localized factors in HRGC safety analysis.