An empirical mode decomposition based recurrent Hermite neural network (ERHNN) prediction model is proposed to predict short-term traffic flow in this study. First, a recurrent Hermite neural network (RHNN) prediction model with different orthonormal Hermite polynomial basis functions (OHPBFs) as activation functions is introduced. Then, to further mitigate the influence of noise and improve the accuracy of prediction, an empirical mode decomposition (EMD) method is derived to decompose the original short-term traffic flow data into several intrinsic mode functions (IMFs) and adopt them as the inputs for the RHNNs. Therefore, an ERHNN prediction model, which comprises good predictive ability for the nonlinear and non-stationary signals through the combination of the merits of OHPBFs, EMD and EHNN, is proposed to predict short-term traffic flow more effectively. The validity of the ERHNN prediction model is verified using all day short-term traffic flow data at high way I–80W in California. Simulation results demonstrate that the proposed ERHNN prediction model is with superior performance compared with the pure recurrent neural network (RNN) and RHNN prediction models.
Short-term traffic flow prediction using EMD-based recurrent Hermite neural network approach
2012-09-01
736219 byte
Conference paper
Electronic Resource
English
Short-term traffic flow prediction method based on graph convolution recurrent neural network
European Patent Office | 2024
|