A hybrid method for short-term freeway travel time prediction based on wavelet neural network and Markov chain

Hang Yang, Yajie Zou, Zhongyu Wang, Bing Wu
2018 Canadian journal of civil engineering (Print)  
1 Short-term travel time prediction is an essential input to Intelligent Transportation Systems 2 (ITS). Timely and accurate traffic forecasting is necessary for Advanced Traffic Management 3 Systems (ATMS) and Advanced Traveler Information Systems (ATIS). Despite several short-4 term travel time prediction approaches have been proposed in the past decade, especially for 5 hybrid models which consist of machine learning models and statistical models, few studies 6 focus on the over-fitting
more » ... em brought by hybrid models. The over-fitting problem 7 deteriorates the prediction accuracy especially during peak hours. This paper proposes a hybrid 8 model which embraces Wavelet Neural Network, Markov Chain and the volatility model (WNN-9 MAR-VOA) for short-term travel time prediction in a freeway system. The purpose of this paper 10 is to provide deeper insights into underlining dynamic traffic patterns and to improve the 11 prediction accuracy and robustness. This method takes periodical analysis, error correction and 12 noise extraction into consideration and improve the forecasting performance in peak hours. The 13 proposed methodology predicts travel time by decomposing travel time data into three 14 components: a periodic trend presented by a modified Wavelet Neural Network (WNN), a 15 residual part modeled by Markov Chain, and the volatility part estimated by the modified 16 generalized autoregressive conditional heteroscedasticity (GJR-GARCH) model. Forecasting 17 performance is investigated with freeway travel time data from Houston, Texas and examined by 18 three measures: mean absolute error (MAE), mean absolute percentage error (MAPE), and root 19 mean square error (RMSE). The results show that the travel times predicted by the WNN-MAR-20 VOA method are robust and accurate. Meanwhile, the proposed method is able to capture the 21 underlying periodic characteristics and volatility nature of travel time data. 22
doi:10.1139/cjce-2017-0231 fatcat:d4vyqtmex5d4nchu7erextkedm