Using conditional inference forests to identify the factors affecting crash severity on arterial corridors

Area of Science:

• Traffic Safety Research
• Transportation Engineering
• Data Mining in Road Safety

Background:

• Urban arterial roads experience a significant number of fatal and severe crashes.
• Identifying contributing factors to crash severity is crucial for developing effective safety strategies.

Purpose of the Study:

• To identify traffic, highway design, and driver-vehicle factors associated with fatal/severe crashes on urban arterials.
• To analyze crash data using an information discovery approach for observational data.

Main Methods:

• Application of conditional inference forests (a type of Random Forest) to analyze crash data.
• Utilizing chi-square test statistics to measure variable associations and identify significant predictors.
• Employing a novel 'element' variable to categorize crash locations (segments, intersections, access points).

Main Results:

• Alcohol/drug use and higher speed limits are significant contributors to severe crashes.
• Non-use of safety equipment and vulnerable age groups (under 3 or over 55) increase injury severity.
• Segments and access points are identified as riskier locations for single-vehicle crashes.

Conclusions:

• Roadway locations like segments and access points are critical areas for severe crash occurrence.
• Factors such as higher skid resistance and k-factor correlate with increased injury severity.
• The study provides actionable insights for targeted safety improvements on urban arterials.