An application of modified Smeed, adapted Andreassen and artificial neural network accident models to three metropolitan cities of Turkey

Abstract

Traffic accident prediction models are used for transportation safety studies. This study proposes 2 analytical models and an artificial neural network (ANN) model to estimate the number of deaths resulted from traffic accidents in 3 metropolitan cities of Turkey utilizing historical data between 1986 and 2005. The first analytical model is a modified form of the Smeed accident prediction model and the second one is an adapted form of the Andreassen model. In the model development, population (P) and the number of vehicles (N) were taken as independent variables. In the ANN approach, the sigmoid and pureline functions were used as activation functions with feed forward-back propagation algorithm. The model estimations were compared against the observations and it was seen that the ANN model performed better than the other 2 analytical models. Key words: Accident prediction model, the Smeed model, the Andreassen model, traffic accident deaths, artificial neural network (ANN), Turkey.