Study of Traffic Flow Short-Time Prediction Based on Wavelet Neural Network

Abstract

In order to improve the accuracy of traffic flow short-time prediction, a traffic flow short-time forecasting model is presented based on wavelet neural network. Firstly, a traffic flow short-time prediction model is established based on improved BP neural network. Secondly, a wavelet neural network prediction model is created to improve slow convergence speed and low forecasting precision of BP neural network. The excitation function of hidden layer use wavelet function instead of sigmoid function. Wavelet neural network combines local characteristics of wavelet transform with self-learning capability of neural network. So it has strong approximation and tolerance. Finally, the two models are used to solve traffic flow short-time prediction separately; simulation results show that wavelet neural network is better than BP neural network. Wavelet neural network has high convergence speed and forecasting precision.