Traffic incident management and estimation of duration: Case of Istanbul Trans European Motorway (TEM)

Unexpected events such as traffic crashes, disabled vehicles, flat tires, spilled loads, empty fuel etc. works cause the traffic congestion or increase the level of the existing traffic congestion on the roadway. Solution or reduce the effects of this problem would be possible with incident management that delivers planned, systematic and a well-coordinated human being, institutional, and technical resource within. Traffic incident management requires estimation incident duration from the start to the final stage of the incident for the identification of the strategies. Hence, the negative effects of the incidents have been minimized with managing the process that between the incidents occurred and the traffic returned to normal in an efficient manner. For this purpose the scope of the thesis, traffic accidents data was obtained that occurred on the Istanbul TEM. This accident data are divided into three groups according to duration with the K-means clustering. With using the accident data and the group's properties, CHAID and C&RT decision tree time estimation models have been created and have been tested. Test results show that the prediction accuracy was above of 76% for both models. According to these results, it was seen that both models generated are usable.