Incident Duration Model on Urban Freeways Using Three Different Algorithms of Decision Tree

Effective incident management requires accurate prediction of incident duration. In this paper, Classification and Regression Tree (CART), CHAID and Exhaustive CHAID is employed to model the incident duration. All 65000 incident records from Beijing Transportation Management Bureau are used for model establishment and another 8000 records for validation. The average relative error of the CART model is 29.5197% while CHAID is 30.78%; Exhaustive CHAID is 31.23%. It shows that the reliability of the three models is quite satisfactory. The average relative error of the prediction on different ring roads of Beijing is approximately the same.