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

Area of Science:

• Road Safety Engineering
• Transportation Science
• Machine Learning Applications

Background:

• Child pedestrian crashes pose a critical road safety concern, frequently leading to severe or fatal injuries.
• Limited research comprehensively analyzes the contributing factors to child pedestrian crash severity.

Purpose of the Study:

• To identify and explain major risk factors for child pedestrian crash severity.
• To generate evidence-based safety recommendations using advanced analytical techniques.

Main Methods:

• Hybrid approach using machine learning (XGBoost, Random Forest) and Cluster Correspondence Analysis (CCA).
• SHAP analysis for explainable AI insights into factor impacts.
• Utilized Texas crash data (2017-2022) involving 4,762 child pedestrian crashes.

Main Results:

• Crashes frequently occur on city streets, intersections, and non-trafficway areas.
• Poor lighting, high speeds, and lack of traffic control in low-speed zones increase severity.
• Afternoon/evening urban and nighttime rural crashes are riskier; turning movements exacerbate risks.

Conclusions:

• Child pedestrian safety is influenced by infrastructure, lighting, speed, and environmental factors.
• Targeted interventions are needed to mitigate risks associated with specific road environments and times.
• Improved infrastructure, lighting, and speed management are crucial for child pedestrian safety.