Accident prediction models for roads with minor junctions

Area of Science:

• Road safety engineering
• Traffic accident analysis
• Statistical modeling

Background:

• Accurate prediction of road accidents is crucial for effective safety interventions.
• Existing methods often struggle with data limitations, particularly the absence of minor approach traffic counts.
• Understanding accident risk factors on main roads with numerous minor junctions is essential.

Purpose of the Study:

• To develop and validate a novel method for predicting expected accidents on main roads with minor junctions.
• To address the challenge of unavailable traffic counts on minor road approaches.
• To provide a reliable tool for road safety assessment in diverse highway environments.

Main Methods:

• Utilized generalized linear modeling (GLM) to develop regression estimates for accident prediction.
• Incorporated an empirical Bayes (EB) procedure to refine estimates by integrating accident data.
• Analyzed a comprehensive dataset of approximately 3800 km of UK highways, including over 5000 minor junctions.

Main Results:

• Accidents on highway sections demonstrated a non-linear relationship with exposure and minor junction frequency.
• Regression model estimates were found to be superior to simple accident counts for prediction.
• The empirical Bayes method yielded the most accurate and unbiased accident predictions, particularly for high-risk sites.

Conclusions:

• The empirical Bayes method is the preferred approach for estimating expected accidents when minor traffic data is absent.
• This validated method improves the accuracy of accident prediction, aiding in targeted road safety improvements.
• The findings are applicable to both single and dual-carriageway roads in urban and rural settings.