A data-driven approach to child pedestrian crash analysis using dimension reduction, clustering, and explainable AI

Child pedestrian (less than 15 years old) crashes are a critical road safety concern, often resulting in severe or fatal injuries due to the inherent vulnerability of young pedestrians. Despite the pressing nature of this issue, limited studies have comprehensively analyzed the key contributing factors influencing crash severity outcomes. Addressing this gap, the current study utilizes hybrid approach employing machine learning models, such as XGBoost and Random Forest, and Cluster Correspondence Analysis (CCA), a joint dimension reduction and clustering method to identify and explain the major risk factors associated with child pedestrian crash severity, and to generate evidence-based safety recommendations through advanced analytical techniques. SHAP analysis, an explainable AI, was conducted on each cluster to further investigate the impact of factors on crash severity. The study uses crash data from the Texas Department of Transportation's Crash Records Information System (CRIS) spanning 2017 to 2022, including 4,762 crashes involving pedestrians under 15 years of age. Key variables include roadway factors (intersection type, traffic control, posted speed limits, road class), temporal aspects (crash hour, season), demographic attributes (age, ethnicity), and environmental conditions (weather, lighting). The results reveal that child pedestrian-involved crashes are associated with frequent interactions on city streets, intersections, and non-trafficway areas like parking lots and driveways. Poor lighting conditions, particularly in rural and high-speed road environments, and absent traffic controls in low-speed zones significantly contribute to crash severity. Temporal patterns indicate higher risks during afternoon and evening hours in urban areas and nighttime in rural settings, with turning movements further increasing risks due to difficulties in detecting pedestrians. This study provides critical insights into the multifaceted factors influencing child pedestrian crash severity, emphasizing the need for improved infrastructure, enhanced lighting, speed management, and targeted safety interventions.