Multivariate analysis of roadway multi-fatality crashes using association rules mining and rules graph structures: A...

Area of Science:

• Traffic Safety Engineering
• Data Mining
• Transportation Systems Analysis

Background:

• Roadway multi-fatality crashes pose a significant threat to public safety.
• Understanding the complex interplay of factors contributing to these severe events is crucial for effective prevention.

Purpose of the Study:

• To explore contributory factors and interdependent characteristics of multi-fatality crashes.
• To develop and apply a novel framework integrating association rules mining and rules graph structures for crash analysis.

Main Methods:

• Utilized association rule mining to analyze 1068 severe fatal crashes in China (2015-2020).
• Constructed modular rules graph structures using graph theory to visualize factor interactions.
• Generated 1452 significant association rules.

Main Results:

• Identified key factors associated with multi-fatality crashes: improper operations, passenger overload, fewer lanes, mountainous terrain, and run-off-the-road events.
• Demonstrated unique crash patterns and association rules across different severity levels, road types, and terrains.
• Found that 43% of severe crashes involved a combination of human-vehicle-road-environment factors, compared to 3% for normal crashes.

Conclusions:

• Hidden associations between crash factors significantly contribute to the severity and frequency of multi-fatality incidents.
• Crash mechanisms in multi-fatality events are systemically more complex than in typical crashes.
• The proposed framework effectively maps risk factors, offering insights for targeted transportation safety interventions.