Accurate prediction of traffic flow is essential for effective traffic management. This paper proposes a long-short-term memory network ensembled with variational mode decomposition and wavelet denoising (VMD-WD-LSTM). First, the LSTM model is employed as a basic prediction model. Subsequently, variational mode decomposition (VMD) is applied to decompose the time series, followed by a stationarity test on the resulting components. The intrinsic mode functions (IMFs) with non-stationary characteristics are further denoised using wavelet denoising. Finally, the processed time series are input into the input layer of the LSTM model, respectively. The final short-term traffic flow prediction is obtained by reconstructing each time series prediction result. Six sets of traffic flow data are selected to test the model. The results show that the proposed model is superior to other four methods.
Wang et al. (Thu,) studied this question.