Prediction Model for Urban Expressway Short-Term Traffic Flow Based on the Support Vector Regression

Abstract

Accurate forecasting of short-term traffic flow played an important role in Intelligent Transportation Systems (ITS) to prevent or mitigate congestions in metropolitan areas. The prediction model based on the Support Vector Regression (SVR) was built to improve the prediction accuracy of short-term traffic flow of urban expressway. The data pre-processing and model-parameter-selection were discussed. The prediction model was validated by short-term traffic flow data, which was collected at a certain expressways in Beijing. The experiment result showed that the prediction model could achieve the highest accuracy while was equal to 0.25. The predicted data showed very good agreements with the ground-truth data, and the result was satisfactory. On the other hand, the maximum relative error is 0.405% for the prediction model based on SVR. The large errors happened before 7:00 and after 22:00, when traffic were not so heavy. The prediction model is of high precision and quite feasible for applications.