Exploring nighttime pedestrian crash patterns at intersection and segments: Findings from the machine learning algorithm

Abstract

Introduction: Pedestrian safety at nighttime is an ongoing critical traffic safety concern. Although poor visibility is primarily associated with nighttime pedestrian crashes, other contributing factors such as humans, vehicles, roadways, and environmental factors interact with each other to cause a crash. Additionally, the pattern of nighttime pedestrian crashes differs significantly according to the intersection and segment location, which requires further exploration.

Data: This study applied Association Rules Mining (ARM), a rule-based machine learning method, to reveal the association of nighttime pedestrian crash risk factors according to the intersection and segment locations using 2,505 nighttime pedestrian fatal and injury crashes in Louisiana (2015-2019).

Results and conclusions: Based on the generated rules, the results show that nighttime pedestrian crashes at the intersection are associated with right-turn vehicle movement, older drivers (>64 years) at the high-speed intersection, senior pedestrians (>64 years) in rainy weather conditions, violation by pedestrian age group '<15 years', and alcohol-intoxicated pedestrian violation in business/industrial areas. Additionally, 'careless operation' at the intersection is associated with alcohol-involved drivers. Most of the nighttime pedestrian crashes at segments are associated with roadways with no physical separation and the absence of streetlights. Driver alcohol involvement and their physical condition (inattentive/distracted) are also associated with pedestrian crashes associated at the segment location at night. Other segment pedestrian crashes are linked to the interstate in dark conditions, open country locations, and high-speed roadways. Additionally, the crash site investigation identified several critical pedestrian safety concerns including the lack of crosswalk facilities, high driveway density, and pedestrian behavioral patterns (e.g., crossing at roadway segments close to the intersection location).

Practical applications: The findings of this study can be used for selecting the appropriate countermeasures based on a case-by-case basis. The exposure patterns can be used in educational campaigns to strategically reduce nighttime pedestrian crashes.

Keywords: Alcohol; Dark conditions; Fatal; High-speed intersection; Interstate; Machine learning.