Prediction of Freeway Incident Duration based on Classification Tree Analysis

概要

論文の詳細を見る
How to provide accurate travel time under incidents is one of the most important premises for highway safety management. Although there was an increase in the studies of predicting incident duration recently, predicting incident duration accurately is currently still a challenging technology due to the quality of incident dataset. Additionally, the previous studies still have some limitations for empirical implication, such as pre-defined function form and strict statistical assumptions. This study employed the huge incident data, recorded accurately from Taiwan freeway systems, to develop predicting incident duration model without the disadvantages of traditional statistical techniques. The analysis results represented that the most two important variables are number of large-sized vehicles and incident type, and provided decision rules to predict incident duration. Therefore, the motorists will avoid traffic jam by changing travel route while the traffic management organization makes decisions timely to mitigate traffic congestion and clear the incident effectively.