A New Holistic Crashes Prediction Model based on Zero-Truncated data for Intercity Four-Lane Highways Curves

Document Type : Research Paper


1 Ph.D. candidate, Department of Civil Engineering, faculty of civil and earth resources engineering, Centeral Tehran Branch, Islamic Azad University, Tehran, Iran.

2 Professor, Department of Civil and Environmental Engineering, Tarbiat Modares University, Iran

3 Assistant Professor, Science and Research Branch, Islamic Azad University, School of Civil Engineering, Tehran, Iran


This study is aimed at exploring the effect of some recognized and new candidate variables of horizontal curves on crash frequency in four-lane highways using zero-truncated crash data. The present study has considered the related variables for 45 curves of four-lane intercity highways during a three-year period (2018-2020). The standard Poisson distribution is a benchmark for modeling Equi-dispersion count data and could not express Under-dispersion zero-truncated data. The modeling was performed using Poisson, Negative Binomial, Zero-Truncated Poisson, Zero-Truncated Negative Binomial, and Conway-Maxwell Poisson (COM-Poisson) regression. The results revealed that the COM-Poisson regression distribution could effectively fit the model Under-dispersion zero-truncated Crashes data. According to the results, using the consistency and self-explaining variables as a useful approach for the estimation of crash frequency in four-lane highway horizontal curves was evaluated.