| Summary: In this study, the sections, which have nearly 40–50 percent of trucks in the total vehicle traffic, are considered in Ankara-Aksaray-Eregli rural divided highways and the relationship between truck accidents and traffic and highway geometric characteristics is modelled with the statistical and Artificial Neural Networks methods. In the first part of the study, Poisson, Negative Binomial and Zero Inflated Negative Binomial Regression models are used for defining the relationship between truck accidents and traffic and highway geometric characteristics. Maximum likelihood method is used to predict the parameters of these models and deviance, Akaike Information Criteria (AIC) and Vuong values are evaluated as goodness of fit. As conclusion, Negative Binomial Regression model is chosen to be the best statistical model for defining the truck accident involvements. In the second part of the study, Negative Binomial Regression method and Artificial Neural Networks are compared with each other. In all data, Artificial Neural Network model performs slightly better than the Negative Binomial Regression model in terms of the prediction of the least errors. As conclusion, the Artificial Neural Networks is proposed as an alternative method for modeling truck accidents. |
