Incorporating behavioral variables into crash count prediction by severity: A multivariate multiple risk source...

Area of Science:

• Transportation Engineering
• Traffic Safety Research
• Statistical Modeling

Background:

• Traffic crash analysis traditionally uses frequency and severity as indicators of risk.
• Existing models often omit crucial non-engineering factors, particularly driver behavior.
• A comprehensive approach is needed to integrate diverse risk factors for effective safety programs.

Purpose of the Study:

• To develop a robust modeling technique for simultaneous crash frequency and severity prediction.
• To address the limitations of conventional models in accounting for variations between risk sources.
• To explicitly incorporate behavioral factors alongside engineering factors in crash prediction.

Main Methods:

• Development and application of a multivariate multiple risk source regression technique.
• Modeling crash frequency and severity concurrently, accounting for correlations between severity levels.
• Comparison of the proposed model with single-equation negative binomial and univariate multiple risk source models.

Main Results:

• The multivariate multiple risk source model demonstrated superior statistical fit compared to conventional and univariate models.
• The model successfully captured correlations between crash severity levels.
• It effectively identified varying impacts of factors from distinct engineering and behavioral risk sources.

Conclusions:

• The proposed multivariate model offers a significant advancement in traffic crash prediction.
• It provides deeper insights into distinct sources of crash risk by integrating behavioral factors.
• Findings can inform safety practitioners and guide targeted roadway improvement programs.