Review of crash prediction models and their applicability in black spot identification to improve road safety

Review of crash prediction models and their applicability in black spot identification to improve road safety

Objective: This study aims to review of the development of crash prediction models, and their applications to analyse and identify black spots to improve road safety.

Methods: Several modelling techniques have been reviewed in this study including, multiple linear regression, Poisson distribution, negative binomial, random effect technique, and multiple logistic regression models to identify their suitability to develop the crash prediction models. The studies related to the identification of black spots were also reviewed based on the type of crash data used in the identification process.

Result: The reviewed documents highlight the shortcomings within the traditional crash prediction models (CPMs), as well as the demonstrated the flexibilities and effectiveness of the latest methods. Most suitable models can now be developed to represent the actual scenarios from several modelling techniques, where they provide a realistic and accurate prediction of crash frequency, for example, to determine if the location had a traffic safety problem compared to other locations with similar conditions and to identify the suitable measures to reduce crashes.

Application/Improvements: The models identified in this research are already being used but the modelling approaches can be further modified to include the latest technical application on roads, available post-crash management system or safety culture which are commonly related road safety outcomes.

Open Access.