Nighttime Pedestrian Safety in Different Communities: Application of Artificial Intelligence Techniques

Usman, Sheikh M.; Khattak, Asad J.; Patwary, A. Latif · 2025 · ROSA P / University of Tennessee, Knoxville. Department of Civil and Environmetal Engineering

archive: archived pipeline: cataloged verified

Get this paper ↗ (full text — opens at the source; we link to it, we don't host it)

Summary

This study addresses the critical issue of nighttime pedestrian safety in the United States, where pedestrian fatalities have risen significantly, with over 75% occurring at night. Motivated by the USDOT’s Vision Zero goal and the need to understand disparities in transportation-burdened communities, the research investigates the correlates of nighttime pedestrian crash injury severity. It specifically examines how six socioeconomic and environmental indicators—Economy, Health, Language Proficiency, Resilience, Environmental, and Transportation Access—interact with conventional crash factors to influence injury outcomes. The researchers analyzed 2,329 nighttime pedestrian crashes in North Carolina from 2016 to 2019, selected to ensure pre-pandemic data homogeneity. Crash data were extracted from police reports using the Pedestrian and Bicyclist Crash Analysis Tool and integrated with census tract-level indicator data from the USDOT. The methodology employed a two-stage approach: first, an inference-based Ordered Logit model was used to identify significant correlates of injury severity. Second, to improve predictive accuracy for planning purposes, a heterogeneous ensemble method known as “Stacking” was applied. This AI-based framework combined four base learners—Ordered Logit, Decision Tree, Gradient Boosting, and Random Forest—using Random Forest as the meta-learner to aggregate predictions. The results demonstrated that the stacked ensemble model achieved a predictive accuracy of 78.85%, outperforming the best individual base learner, which reached 73.56%. The analysis revealed significant associations between injury severity and several factors, including Economy and Transportation Access indicators, the absence of roadway lighting, pedestrian crossing violations, and alcohol impairment. Descriptive statistics highlighted that nighttime crashes resulted in higher proportions of serious and fatal injuries compared to daytime crashes. Additionally, Black pedestrians were disproportionately overrepresented in nighttime crashes relative to their population share, while crashes involving impaired pedestrians were significantly more frequent at night. Most crashes occurred on local roads without lighting, with pedestrians frequently positioned in travel lanes rather than at crosswalks. The study concludes that integrating socioeconomic indicators with advanced AI modeling provides a robust framework for forecasting pedestrian crash severity. These findings offer practical applications for safety practitioners and policymakers, enabling the implementation of targeted interventions. By identifying specific infrastructural deficits and community burdens, stakeholders can prioritize resources to improve roadway infrastructure, such as lighting and pedestrian crossings, in high-risk areas. This approach supports the Safe Systems Approach by addressing both safe users and safe road infrastructure, ultimately aiming to reduce disparities and enhance overall pedestrian safety in diverse communities.

Key finding

A heterogeneous ensemble stacking model combining ordered logit, decision tree, gradient boosting, and random forest algorithms achieved a predictive accuracy of 78.85% for nighttime pedestrian crash injury severity, outperforming individual base models and identifying socioeconomic indicators, lighting conditions, and impairment as key correlates.

Methodology

modeling

Sample size: 2329

Provenance

The full processing record for this entry. Every stage of this paper's journey through the pipeline is logged — what ran, with which tool and model, how many attempts it took, and when it last completed. Discovered via bulk_ingest_rosap on 2026-05-23 (6 acquisition events logged).

Summary generated by qwen3.6-27b-prismaquant on 2026-06-10; verification: verified.

Topics

Ranked by relevance to this paper. Hover a topic for its definition.

• demographic disparities
• vru crash typology
• incidence prevalence

Information type

What kind of knowledge this paper contributes, grouped by family — independent of topic (what it is about) and method (how it was studied).

• Empirical Findings: crash risk outcomes