STUDY OF SHORT-TERM TRAFFIC FLOW FORECASTING MODELS BASED ON STATISTICS LEARNING THEORY.

Short-term traffic flow prediction is most popular in ITS. Based on the shortcomings of the previous study and the difficulties of traffic data collection, this paper proposed a new short-term traffic flow forecasting model based on the statistics learning theory (SLT). In addition, this study compares Support Vector Machine (SVM) model with the traditional Radial Basis Function Artificial Neural Networks (RBFANN) model and Auto Regressive (AR) time series model. The results based on real data show that SVM forecasting model provides a more promising prediction result than RBFANN and AR model. Moreover, the study discussed the effect on forecasting with denoising process using wavelet transform. For the covering abstract see E134653.