Exploring analytical, simulation-based, and hybrid model structures for multivariate crash frequency modeling

Abstract

In safety literature, there are two ways to incorporate the potential correlation between multiple crash frequency variables: (1) simulation-based approach and (2) analytical closed-form approach. The current research effort undertakes a comparison between simulation-based multivariate model and copula based closed-form approach to analyze zonal level crash counts for different crash types. Further, the research builds on earlier copula based models by incorporating random parameters thus proposing a hybrid (combination of analytical and simulation based system) approach to incorporating unobserved heterogeneity. Within the proposed hybrid copula model, the empirical analysis involves estimation of count models using four different copula structures which cover a wide range of dependency structures, including radial symmetry and asymmetry, and asymptotic tail independence and dependence. Further, to the best of authors’ knowledge, this study is the first of its kind to incorporate attribute variability (random parameters) effect within the copula framework. The empirical analysis is based on traffic analysis zone (TAZ) level crash count data for both motorized and non-motorized crashes from Central Florida for the year 2016. A comprehensive set of exogenous variables including roadway, built environment, land-use, traffic, socio-demographic and spatial spillover characteristics are considered for the analysis. The resulting data fit and prediction performance offered by the proposed approach clearly highlights the hybrid model – Random Parameter Copula based approach’s superiority over the purely simulation-based multivariate model in our study context. The comparison exercise is further augmented by undertaking an in-depth comparison for different count events across different crash types and a correct classification analysis. The estimated results further reinforce the improved performance of the Random Parameter Copula-based multivariate approach. The applicability of the model for hot spot identification is illustrated by generating plots identifying high-crash and low- crash zones by crash type in the Central Florida region.