An analysis of urban collisions using an artificial intelligence model

Area of Science:

• Traffic Safety Engineering
• Urban Planning
• Data Science

Background:

• Traditional road accident studies often focus on estimating variable effects and accident counts.
• Existing models may not fully capture the complex interactions influencing urban intersection safety.

Purpose of the Study:

• To introduce an alternative method for modeling urban vehicular accidents using artificial neural networks (ANN).
• To quantify the degree of danger at urban intersections under various scenarios.
• To analyze the relationship between intersection characteristics and accident risk.

Main Methods:

• A descriptive statistical analysis of road accidents in Milan from 1992-1995.
• Development and application of an artificial neural network (ANN) model for accident analysis.
• Quantification of intersection danger using the ANN model across different scenarios.

Main Results:

• The ANN model provides an innovative methodology for urban vehicular accident modeling.
• Intersection complexity and traffic regulation significantly influence the accident index.
• Non-signalized intersections at night exhibit the highest risk for pedestrian run-over accidents.

Conclusions:

• Artificial neural networks offer a powerful tool for understanding and mitigating urban traffic accident risks.
• Intersection design and traffic management strategies should consider complexity and specific risk factors.
• Targeted interventions are needed for high-risk scenarios, such as non-signalized intersections during nighttime.