Abstract
This research developed an incident detection model by the Bayesian Network approach using traffic detector data measured on Tokyo Metropolitan expressways. Traffic state information including traffic volume, velocity, occupancy and occurance of incidents is used to make conditional probability table that is the output from the Bayesian Network model. In this paper, the model performance is verified by not only detection rates but also detection delays and false rates focusing on five routes on the metropolitan expressways to consider the possibility of installing the proposed model in traffic control center of the expressway. As the results of verification, the detection rate by the proposed model was about 70% in the case of the most accurate route, and the false detection rate was 4%, better results than the existing study that investigates the incident detection model by another method on the same route of the expressway.
