Y. Liu, H.Z. Li

Pages: 69-80

Abstract
In order to solve the problems of low prediction accuracy and high time consumption in traditional traffic flow prediction methods of road construction section, a traffic flow prediction method based on wavelet neural network is proposed. This paper analyzes the characteristics of the road construction section, determines the vehicle operation status and vehicle density of the construction section according to the change of the signal control cycle of the construction section, and completes the extraction of the traffic flow characteristic data of the construction section through the Greenhill linear model; analyzes the basic operation principle of the wavelet neural network, determines the topological structure characteristics of the network, and uses the gradient descent method to process the wavelet neural network . Then, the traffic flow characteristic data is input into the network, and the output value of traffic flow prediction sample data of construction section is normalized by zero mean value standardization method to complete the traffic flow prediction of road construction section. The experimental results show that the accuracy of the proposed method is about 90%, and the longest time cost is about 2.5 s.
Keywords: wavelet neural network; road construction section; traffic flow; greenhill linear model; gradient descent method