Predicting crash risk and identifying crash precursors on Korean expressways using loop detector data.

Area of Science:

• Traffic Engineering
• Road Safety
• Transportation Systems Analysis

Background:

• Crash mechanisms vary by expressway segment type and traffic flow conditions.
• Proactive safety management requires tailored strategies considering these variations.
• Existing models may not adequately address irregular detector spacing common in some expressway networks.

Purpose of the Study:

• To develop real-time crash risk prediction models for expressways.
• To differentiate models based on segment types (basic vs. ramp vicinity) and traffic flow states (uncongested vs. congested).
• To investigate the impact of irregular loop detector intervals on crash risk prediction.

Main Methods:

• Conditional logistic regression was used to identify significant traffic variables, controlling for confounding factors.
• Genetic programming was employed to develop separate crash risk prediction models for distinct conditions.
• The influence of detector station intervals was specifically analyzed.

Main Results:

• Crash risk prediction models demonstrated that traffic flow characteristics leading to crashes differ by segment type and traffic flow state.
• Variables related to detector station intervals significantly influenced crash risk prediction, particularly under uncongested conditions.
• The proposed models exhibited higher prediction performance compared to a single model or traditional logistic models.

Conclusions:

• Tailored, real-time crash risk prediction models are essential for effective proactive safety management on expressways.
• The developed models provide improved performance by considering segment types, traffic flow states, and irregular detector spacing.
• Findings support the implementation of advanced safety strategies for diverse expressway environments.