A framework of boundary collision data aggregation into neighbourhoods

Area of Science:

• Transportation Engineering
• Road Safety Analysis
• Data Science

Background:

• Boundary collisions form a significant portion of motor vehicle incidents.
• Inaccurate aggregation of boundary collisions can compromise road safety analyses and management strategies.
• Existing methods for boundary collision aggregation lack precision.

Purpose of the Study:

• To propose a systematic framework for accurate boundary collision aggregation.
• To enhance the precision and recall of boundary collision identification.
• To develop a more effective method for aggregating boundary collisions into neighborhoods.

Main Methods:

• Utilized entropy-based histogram thresholding to determine boundary zone size and identify boundary collisions.
• Established collision density probability distribution for each neighborhood.
• Developed and applied the collision density ratio (CDR) method for boundary collision aggregation.

Main Results:

• The entropy-based histogram thresholding method demonstrated high precision and recall in identifying boundary collisions.
• The proposed CDR method proved more effective than existing half-to-half and one-to-one ratio methods.
• The framework was successfully applied to collision data in the City of Edmonton.

Conclusions:

• The proposed framework offers a systematic and effective approach to boundary collision aggregation.
• Accurate boundary collision data is essential for improving road safety management.
• The CDR method represents a significant advancement in spatial analysis of traffic incidents.