Incident Duration Predication Based on Traffic Boardcasting Information

摘要

Prediction of incident duration not only can give the traffic managers decisive support, such as releasing traffic information in advance, but also can make the traffic participants arrange travel routes and improve travel efficiency. In order to improve the low prediction accuracy and complexity of the model proposed previously, traffic broadcasting information, including incident broadcasting information from the Traffic Management Bureau, and the real-time traffic data from Traffic Information Committee, has been applied to this paper. The duration of traffic incidents were divided into three categories with 3782 incidents. The artificial neural network with rational structure was applied that input variables combines the characteristics of the incident with the traffic data. The result shows the prediction accuracy is satisfactory with the rational structure and parameters. The traffic data is salutary to predict the duration.