An accident prediction model for divided highways: A case study of Trabzon coastal divided highway


Urban Transport 2013, 19th International Conference on Urban Transport and the Environment, Kos, Greece, 29 - 31 May 2013, pp.711-719

  • Publication Type: Conference Paper / Full Text
  • City: Kos
  • Country: Greece
  • Page Numbers: pp.711-719
  • Karadeniz Technical University Affiliated: Yes


Traffic accidents that occurred in Turkey have been increased for every year. Although serious property damages and fatalities are occurred in accidents, Turkey hasn’t been sufficiently overcome this problem.

The main objective for this study is to investigate the factors which cause accidents and to create an accident prediction model which includes relationships between these factors. With this model, the expected number of accidents at divided highways can be predicted and suitable measures for providing road safety can be defined.

For this study, 5 years (2002, 2003, 2004, 2006 and 2007 years) accident data of 113.5 km. road sections of Trabzon coastal divided highway, traffic and highway characteristics of these sections were collected, then an accident prediction model was formed. The technique of generalized linear models (GLMs) was applied to the data. Because of over dispersion of Poisson regression model, Negative Binomial regression model was found the most appropriate approach for the analysis of these data. This model indicates that the vehicle kilometers of travel, the number of pedestrian crossing and average posted speed are significant variables on traffic accident occurrence.