Travel Time Forecasting Based on Phase Space Reconstruction and SVM

Travel time forecasting is an important content of dynamic traffic navigation. Dynamic traffic data collection is the precondition of forecasting. Many traffic data collection methods have been adopted, such as loop inductive vehicle detector, radar detector, video detector, GPS floating car and so on. Due to the widely distribution, GPS floating car has become the most efficient mean to collect instantaneous traffic information. How to take use of the collected GPS floating car data to forecasting the travel time is a popular research topic. In this paper, we develop a travel time forecasting method with the combining phase space reconstruction theory and SVM. Phase space reconstruction theory is used to determine the number of forecasting variable. SVM is used to forecast the future travel time value. The efficiency of the method is illustrated through analyzing Jinan urban traffic data.