Abstract

Averagely, over 1.3 million traffic accidents with deaths, wounds and property losses occur in a year in Turkey. In the last 30 years, 140000 people have died in traffic accidents. The financial losses of the accidents in the last ten years are over 100 billion dollars. In this study, monthly accidents occurring in Turkey were investigated with Artificial Neural Network (ANN) method. With monthly accident data occurring between the years 1994 to 2007, seasonal fluctuations in the months to come and the number of accidents were predicted. Therefore, an attention is attracted to accidents problems, more reliable analyses about the number of traffic accidents are made and it is emphasized that precautions are increased to prevent the accidents. What sort of the accidents trend will follow in the next months was cleared in the ANN model. The best model was chosen according to the Akaike Information Criteria (AIC), Mean Squared Error and the square of correlation of coefficient (R²). The reliability of the model was also tested by statistical analyses on the error terms.

Key words: Traffic accident analyses, Artificial Neural Network, statistical analyses.