Investigating Influence Factors on Traffic Safety Based on a Hybrid Approach: Taking Pedestrians as an Example

Area of Science:

• Transportation Science
• Public Safety
• Data Science

Background:

• Road traffic safety is a critical global concern, with pedestrian safety being paramount.
• Advancements in intelligent transportation systems (ITS) necessitate effective strategies to prevent pedestrian fatalities.

Purpose of the Study:

• To investigate factors influencing pedestrian injury severity in vehicle-involved crashes.
• To propose and validate a hybrid modeling approach for analyzing crash data.

Main Methods:

• Utilized a hybrid approach combining a random parameters logit model and random forest with SHAP (SHapley Additive exPlanations) method.
• Analyzed pedestrian-involved crash data from multiple sensor-based sources.

Main Results:

• Identified eight significant factors influencing pedestrian injury severity.
• The hybrid model demonstrated complementary strengths in factor identification and contribution analysis.
• Certain factors like 'Workday or not', 'Signal control mode', and 'Road safety attribute' were found not to be significant.

Conclusions:

• The hybrid approach provides robust insights into pedestrian-vehicle crash dynamics.
• Findings can inform the development of targeted ITS strategies to improve pedestrian safety.
• A comprehensive analysis considering both significant and important factors is crucial for effective safety interventions.