Data-Driven Techniques for Identifying Factors Affecting the Severity of Driver Injuries in Highway-Railway Grade Crossing Accidents: A Comparative Analysis Using Random Forest, XGBoost,and  Multinomial Logistic Regression

Mousavi, Rayehe sadat; Bagherian rostami, Behnam; Sanagostar, Zahrs sadst; Zayandeh roodi, mohammad ali; Saghian, Zahra; Bagheri, Morteza

doi:10.22068/ijrare.357

Volume 12, Issue 1 (1-2025) IJRARE 2025, 12(1): 9-23 | Back to browse issues page

‎ 10.22068/ijrare.357

Mendeley

Zotero

RefWorks

Mousavi R S, Bagherian rostami B, Sanagostar Z S, Zayandeh roodi M A, Saghian Z, Bagheri M. Data-Driven Techniques for Identifying Factors Affecting the Severity of Driver Injuries in Highway-Railway Grade Crossing Accidents: A Comparative Analysis Using Random Forest, XGBoost,and Multinomial Logistic Regression. IJRARE 2025; 12 (1) :9-23
URL: http://ijrare.iust.ac.ir/article-1-357-en.html

Data-Driven Techniques for Identifying Factors Affecting the Severity of Driver Injuries in Highway-Railway Grade Crossing Accidents: A Comparative Analysis Using Random Forest, XGBoost,and Multinomial Logistic Regression

Rayehe sadat Mousavi

, Behnam Bagherian rostami

, Zahrs sadst Sanagostar

, Mohammad ali Zayandeh roodi

, Zahra Saghian

, Morteza Bagheri

School of Railway Engineering, Iran University of Science and Technology, Tehran, Iran

Abstract: (45 Views)

This study investigates the factors influencing the severity of accidents at highway-rail grade crossings in the United States and explores strategies to mitigate the risks to road vehicle drivers. Two approaches are employed for modeling accident severity: statistical methods, such as multinomial logistic regression, and machine learning techniques, including Extreme Gradient Boosting (XGBoost) and random forest algorithms. The analysis is based on data from the Federal Railroad Administration’s database, covering a twelve-year period (2010-2022). The results identify several key controllable factors that significantly impact accident severity, including vehicle speed, the position of road users, visibility obstructions, the number of cars in the train, and the speed of the train. Among the models tested, XGBoost demonstrated superior accuracy in predicting accident severity compared to multinomial logistic regression and random forest. Based on the findings, several recommendations are proposed to reduce accident risk at grade crossings, such as lowering train speeds, implementing advanced speed control systems, enhancing lighting at crossings, improving barrier inspections, and optimizing train scheduling. These measures aim to enhance safety and minimize collision severity at highway-rail crossings.

Keywords: severity of the incidents, grade crossings between roads and railways, XGBoost, random forest, multinomial logistic regression

Full-Text [PDF 254 kb] (23 Downloads)

Type of Study: Research | Subject: Railway Transportation