Prediction of Lane Clearance Time of Freeway Incidents Using the M5P Tree Algorithm

Abstract

A number of existing studies have attempted to predict freeway incident duration or incident clearance time. Because lane blockage is the main cause of congestion during freeway incidents, it is more beneficial to predict the lane clearance time instead of the incident clearance time for incidents that involve lane blockages. However, previous studies have not developed prediction models for the lane clearance time. This paper utilizes the M5P tree algorithm for lane clearance time prediction, which has advantages, compared with traditional prediction algorithms. These advantages include the M5P tree algorithm's ability to deal with categorical and continuous variables and variables with missing values. The developed model shows that there are a number of variables that affect the lane clearance time, including the number of lanes blocked, time of day, types and number of vehicles involved, the response by the Severe Incident Response Vehicle (SIRV), and traffic management center response and verification times. Comparison results show that the developed model can generally achieve better prediction results than the traditional regression and decision tree models.