Road Crash Prediction Models: A Review of Methods and Applications

Abstract

Road traffic crashes are still the major road safety problems in the world causing for the death of more than 1 million people each year, although the problem is more serious in low- and middle-income countries. Therefore, road crash prediction models play an important role in road safety management to determining both the predicted crash frequency and the contributing factors that could then be addressed by transport policies. Many types of statistical crash prediction models have been proposed for estimating the predicted crash frequencies in road networks, ranging from basic Poisson and negative binomial models to more complicated models, such as zero-inflated and Conway-Maxwell Poisson regression models. However, little effort has been made in assessing the performance and practical implications of these models when they are used for identifying black spot locations. The study aims to critically summarize the global experience on the development and application of CPMs to analyze and identify black spots for road safety improvements. To achieve the objective several crash modelling techniques have been reviewed. The study also reviewed data and methodological issues in the development of crash prediction models including data collection methods, network segmentation and selection of explanatory variables and the application of crash prediction models in black spot identification for road safety improvements. The study identified the limitations of the most traditional crash modelling techniques and examined the flexibilities and effectiveness of the latest crash modelling techniques.