Assessing the Role of Intersection Proximity in Pedestrian Crashes: Insights from Data Mining Approach

Problem
This preprint addresses the gap in understanding pedestrian crash dynamics at non-intersection locations, despite the known complexity of intersections in roadway networks. The disproportionate frequency of pedestrian crashes occurring away from intersections necessitates a focused investigation into the factors influencing these incidents. The authors leverage a crash database from Louisiana State (2017-2021) to explore how proximity to intersections affects crash patterns, aiming to provide actionable insights for traffic safety improvements.

Method
The study employs a novel framework termed “distance to intersection” to categorize pedestrian crashes based on their proximity to intersections. The dataset comprises 3,135 pedestrian crashes, which are divided into three distinct zones: D1 (within 150 ft of an intersection, 1,277 crashes), D2 (151 ft to 435 ft, 1,060 crashes), and D3 (beyond 435 ft, 798 crashes). To analyze the interactions among various contributing factors, the authors utilize Association Rules Mining, focusing on the top 60 association rules (20 per zone) based on lift and support metrics. Additionally, they explore 124 rules using the Lift Increase Criterion (LIC) to identify significant patterns in crash involvement.

Results
The analysis reveals that approximately 50% of non-intersection pedestrian crashes occur within 198 ft of an intersection, underscoring the critical influence of proximity on crash risk. The study identifies specific patterns and associations within each zone, providing a nuanced understanding of the factors contributing to pedestrian crashes. While exact effect sizes and comparative metrics against baseline models are not disclosed, the identification of 60 key association rules indicates a robust framework for understanding crash dynamics.

Limitations
The authors acknowledge several limitations, including the potential for unobserved confounding variables that may influence crash outcomes. The reliance on historical crash data may not account for changes in traffic patterns or pedestrian behavior over time. Additionally, the study is geographically constrained to Louisiana, which may limit the generalizability of the findings to other regions with different traffic dynamics. The authors do not address the potential biases in data collection or the implications of unreported crashes.

Why it matters
This research has significant implications for traffic safety policy and urban planning. By elucidating the relationship between intersection proximity and pedestrian crash patterns, the findings can inform targeted interventions and countermeasures aimed at reducing pedestrian injuries and fatalities. The insights gained from the association rules can guide the development of more effective safety measures, such as improved signage, enhanced crosswalk visibility, and strategic urban design modifications. This work lays the groundwork for future studies to explore pedestrian safety in various contexts and contributes to the broader discourse on traffic safety and urban mobility.

Authors: Ahmed Hossain, Xiaoduan Sun, Subasish Das
Source: arXiv:2604.28065
URL: https://arxiv.org/abs/2604.28065v1