AN ARTIFICIAL NEURAL NETWORK MODEL FOR HIGHWAY ACCIDENT PREDICTION: A CASE STUDY OF ERZURUM, TURKEY

This study presents an accident prediction model of Erzurum's Highways in Turkey using artificial neural network (ANN) approaches. There are many ANN models for predicting the number of accidents on highways that were developed using 8 years with 7,780 complete accident reports of historical data (2005-2012). The best ANN model was chosen for this task and the model parameters included years, highway sections, section length (km), annual average daily traffic (AADT), the degree of horizontal curvature, the degree of vertical curvature, traffic accidents with heavy vehicles (percentage), and traffic accidents that occurred in summer (percentage). In the ANN model development, the sigmoid activation function was employed with Levenberg-Marquardt algorithm. The performance of the developed ANN model was evaluated by mean square error (MSE), the root mean square error (RMSE), and the coefficient of determination (R-2). The model results indicate that the degree of vertical curvature is the most important parameter that affects the number of accidents on highways.