Modelling for identifying accident-prone spots: Bayesian approach with a Poisson mixture model

Abstract

In traditional identification of hot spots, often known as the sites with black spots or accident-prone locations, methodologies are developed based on the total number of accidents. These criteria provide no consideration of whether the accidents were caused or could be averted by road improvements. These traditional methods result in misidentification of locations that are not truly hazardous from a road safety authority perspective and consequently may lead to a misapplication of safety improvement funding. We consider a mixture of the zero-inflated Poisson and the Poisson regression models to analyze zero-inflated data sets drawn from traffic accident studies. Based on the membership probabilities, observations are well separated into two clusters. One is the ZIP cluster; the other is the standard Poisson cluster. A simulation study and real data analysis are performed to demonstrate model fitting performances of the proposed model. The Bayes factor and the Bayesian information criterion are used to compare the proposed model with several competing models. Ultimately, this model could detect accident-prone spots.