Short-term Traffic Flow Forecasting By Support Vector Regression And System Implementation

The research is supported by the Fundamental Research Funds for the central universities (N120804003) called "the Research on the Key Technologies of Real-time Traffic Information Cloud Computing System with Big Data". The prediction of short-term traffic flow is the key part of intelligent transportation control and guidance. The accurate prediction can improve the performance of intelligent transportation and is helpful to crowded cities.Due to the strong nonlinear, stochastic, time_varying characteristics of urban transportation system, the traditional forecasting methods aren’t suable to short-term traffic flow prediction so that the forecasting methods based on artificial intelligence are focused on by more and more researchers recently. Support Vector Machine Regression (SVR) as a machine learning method based on statistical learning theory (SLT) can efficiently solve small samples, nonlinear, high dimensions and local minima problems.Firstly, the current research progresses are summarized for comparing the advantages and disadvantages of forecasting methods on short-term traffic flow and analyzing the main factors on traffic flow. The SVR method is proposed to address the short-term traffic flow prediction, and the problem description, the model formulation, and the algorithm are provided orderly.The key part of this paper is to obtain the reasonable and optimal parameters for the SVR model. Firstly, the trial and error method is proposed to get the ranges of parameters(C, ε, r). The experiment results show that the parameters are acceptable. Moreover, the comparison with the CHDBP and AR methods are done to illustrate the advantage of SVR. Secondly, the Genetic Algorithm is designed to obtain the optimal parameters of SVR. The actual data are obtained to evaluate the performance of the method. The results show that the accuracy is improved by GA_SVR.Finally, the function analysis is developed for the short-term traffic flow prediction system. The prototype system is implemented based on the system design. The whole research will be helpful to improve the efficiency and effectiveness of intelligent transportation system.