Design of Real-time Traffic Information Prediction and Simulation System Based on AOSVR and On-line Learning

Acquiring real-time information about traffic information is one of the important steps toward the realization of ITS. In this paper, we design a real-time traffic information prediction and simulation system with warm start by integrating an accurate on-line support vector regression (AOSVR) with an on-line learning algorithm that used for improving computational rate. And a fluid simulation model is used to simulation the prediction result. The forecasting implementation has showed that the proposed model is faster and more exact than AOSVR when it is applied to an actual real-time forecasting scheme. And the simulation can intuitively show the change of the traffic status, which helps the users to make effective measures.