Accident models of trumpet interchange S-type ramps using by Poisson, negative binomial regression and ZAM

Abstract

This study deals with the traffic accident at the S-ramps of trumpet interchanges. The purposes of the study are to analyze the relations between the driving conditions of ramps and traffic accidents, and to develop the accident models. In pursuing the above, 142 trumpet interchange (AO, AU, BO and BU) are classified into 4 types; the characteristic data of ramp alignments are investigated; the expressway traffic accident records (the 720 case of traffic accidents from September 1, 1998. to December 31, 2003) are collected; and the data of expressway traffic volume are calculated (the average annual daily traffic of each ramp is calibrated by the expressway origin destination survey). In this study, the accident models, using by Poisson, Negative Binomial (NB) and Zero-Altered Model (ZAM) which are all statistically significant were developed. The likelihood ratio of Poisson model is 0.1212 and that of Zero Inflated Poisson (ZIP) model is 0.2020. Also, the analysis shows that the likelihood ratio of Negative binominal regression model is 0.1918 and that of Zero Inflated Negative Binomial (ZINB) model is 0.1993. In summary, it is analyzed that NB model is the greatest among the above 4 models on the trumpet interchange S-type ramps.