An optimal network screening method of hotspot identification for highway crashes with dynamic site length

Application of naïve Bayesian approach in detecting reproducible fatal collision locations on freeway

Detecting high-collision-concentration locations based solely on collision frequency may produce different results compared to those considering the severities of the collisions. In particular, it can lead government agencies focusing sites with a high collision frequency while neglecting those with a lower collision frequency but a higher percentage of injury and fatal collisions. This study developed systematic ways of detecting reproducible fatal collision locations (R) using the naïve Bayes approach and a continuous risk profile (CRP) that estimates the true collision risk by filtering out random noise in the data. The posterior probability of fatal collisions being reproducible at a location is estimated by the relationship between the spatial distribution of fatal-collision locations (i.e., likelihood) and the CRP (i.e., prior probability). The proposed method can be used to detect sites with the highest proxy measure of the posterior probability (PMP) of observing R. An empirical evaluation using 5-year traffic collision data from six routes in California shows that detecting R based on the PMP outperform those based on the SPF-based approaches or random selection, regardless of various conditions and parameters of the proposed method. This method only requires traffic collision and annual traffic volume data to estimate PMP that prioritize sites being R and the PMPs can be compared across multiple routes. Therefore, it helps government agencies prioritizing sites of multiple routes where the number of fatal collisions can be reduced, thus help them to save lives with limited resources of data collection.

A new spatial count data model with Bayesian additive regression trees for accident hot spot identification

The identification of accident hot spots is a central task of road safety management. Bayesian count data models have emerged as the workhorse method for producing probabilistic rankings of hazardous sites in road networks. Typically, these methods assume simple linear link function specifications, which, however, limit the predictive power of a model. Furthermore, extensive specification searches are precluded by complex model structures arising from the need to account for unobserved heterogeneity and spatial correlations. Modern machine learning (ML) methods offer ways to automate the specification of the link function. However, these methods do not capture estimation uncertainty, and it is also difficult to incorporate spatial correlations. In light of these gaps in the literature, this paper proposes a new spatial negative binomial model which uses Bayesian additive regression trees to endogenously select the specification of the link function. Posterior inference in the proposed model is made feasible with the help of the Pólya-Gamma data augmentation technique. We test the performance of this new model on a crash count data set from a metropolitan highway network. The empirical results show that the proposed model performs at least as well as a baseline spatial count data model with random parameters in terms of goodness of fit and site ranking ability.