Hybrid Methods for Short-Term Traffic Flow Prediction Based on ARIMA-GARCH Model and Wavelet Neural Network

Accurate short-term traffic flow prediction is essential for real-time traffic control. A linear hybrid method and a nonlinear hybrid method for short-term traffic flow prediction are proposed with vehicle type as one concern. Traffic flow data are divided into the similar, volatile, and irregular parts. The selected methods are the autoregressive integrated moving average and generalized autoregressive conditional heteroscedasticity (ARIMA-GARCH) model, the Markov model with state membership degree, and the wavelet neural network. The ARIMA-GARCH model is used to predict the similar and volatile parts, and the other methods are adopted to predict the irregular part. This paper aims at providing better prediction methods for short-term traffic flow, and comparing the advantages and disadvantages of the linear and nonlinear hybrid methods. Additionally, the impacts of vehicle type on the predicted values are analyzed. The proposed methods are tested using field data from Dalian, China, and Hefei, China. The results suggest that the developed nonlinear hybrid method should be used with vehicle type and sampling interval as concerns.