Short-term traffic flow prediction based on secondary hybrid decomposition and deep echo state networks

Short-term traffic flow prediction is a significant and challenging research topic as it is closely related to the application of intelligent transportation systems. Due to the variable and random characteristics of the transportation system, raw traffic flow data often contain noise, and predicting the raw data directly may reduce the accuracy and effectiveness of the prediction models. Therefore, a hybrid method is established in this research which combines denoising schemes and deep learning models to improve the prediction accuracy. The time series denoising schemes include two parts: the complete ensemble empirical mode decomposition with adaptive noise (CEEMDAN), and wavelet packet decomposition (WPD). Firstly, the raw traffic flow data are decomposed by CEEMDAN to obtain intrinsic mode functions (IMFs) and a residual. Then the IMFs are divided into anti-persistent and persistent components through the Hurst Exponent index. The anti-persistent components are re-decomposed by the WPD algorithm, and persistent components are aggregated into one component. Finally, these components and residual are forecasted by the deep echo state network (DeepESN) model. In the experiment, to investigate the prediction performance of the proposed CEEMDAN-WPD123456–7a11-DeepESN model, the LSTM, CEEMDAN-LSTM, CEEMDAN-WPD-LSTM, DeepESN, CEEMDAN-DeepESN, CEEMDAN-WPD1-DeepESN, CEEMDAN-WPD123456-DeepESN and CEEMDAN-WPD1a6–7a11-DeepESN models are considered to be comparison models. The experimental results demonstrate that the proposed model has superior performance on both efficiency and accuracy.