Degree

Master of Science in Engineering (MSE)

Department

Civil Engineering

Document Type

Thesis

Abstract

Speeding is a major contributing factor to roadway crashes and remains a significant safety concern. Understanding the factors influencing speeding crash severity and how they vary across roadway environments is essential for developing effective safety countermeasures. A total of 22,151 speeding-related crashes recorded in Louisiana between 2015 and 2024 were analyzed. Crashes were categorized by FHWA functional classification into three roadway groups: Interstates and Principal Arterials, Minor Arterials and Collectors, and Local Roads. Four spatial machine learning models, GXGBoost, SDT, GRF, and GeoKNN, were evaluated to classify crash severity outcomes, as traditional models do not account for the spatial dependence and geographic heterogeneity present in crash data. To improve model interpretability, SHAP was applied to identify the most influential predictors of crash severity. Among the evaluated models, GXGBoost achieved the highest classification performance, producing peak F1-scores of 69.7% for interstates and principal arterials, 66.2% for minor arterials and collectors, and 67.0% for local roads. The results also revealed differences in spatial structure across roadway types. SHAP analysis identified several factors consistently associated with crash severity outcomes. Ambulance presence and clearance times exceeding two hours were associated with injury and fatal crashes. Motorcycle involvement increased fatal crash likelihood, while drug presence was the strongest predictor of fatal crashes on minor arterials and collectors. Wet surface conditions increased no-injury likelihood on minor arterials and collectors and local roads but decreased it on interstates and principal arterials. On local roads, weekend conditions, off-roadway crashes, posted speed limit (40-45mph), and police response times exceeding 15 minutes were associated with increased crash severity. The results demonstrate that speeding-related crash severity on Louisiana's highways is shaped by spatial context, roadway functional classification, driver behavior, vehicle type, environmental conditions, traffic characteristics, and emergency response factors.

Date

2-6-2026

DOI

https://proquest.com/docview/3347830246

First Committee Chair

Elisabeta Mitran

Second Committee Chair

Vijaya Gopu

First Committee Member

M. Rahman

Second Committee Member

Milhan Moomen

Third Committee Member

Mohammed Khattak

Download

Share

COinS