Pedestrian crash patterns at signalized and unsignalized intersections: Insights from cluster correspondence analysis and association rule mining

Pedestrian safety at intersections remains a critical concern in urban transportation planning. This study aims to identify broad pedestrian crash patterns through CCA-based attribute clustering and determine the factors linked to crash severity, providing insights that can help inform the development of more effective safety measures. This study employs unsupervised machine learning to analyze crash patterns and identify factors contributing to pedestrian crashes at signalized and unsignalized intersections. Using a 5-year dataset (1874 pedestrian crashes, 2015–2019) from the Louisiana Department of Transportation and Development, Cluster Correspondence Analysis (CCA), a statistical technique combining dimension reduction and clustering, identified three broad crash patterns per intersection type. Daytime crashes were linked to driver distraction, while nighttime crashes involved impaired pedestrians in dark conditions. Association Rule Mining (ARM), exploring associations with severity, revealed that signalized intersections with speed limits ≥50 mph in dark conditions were more than twice as likely to result in fatal or severe injuries. These findings provide insights for targeted interventions and policies to enhance pedestrian safety.