Wavelet Neural Network Algorithm And The Application In Macroscopic Dynamic Characteristics Of Traffic Flow

With the continuous progress of science and technology, the level of the nationalindustrialization improves constantly, which greatly promotes the development ofeconomic. The automotive industry gets a lot of opportunities in this process. However,this development has brought lots of issues to the society, such as energy consumption,air pollution, traffic congestion and other social problems, among which the trafficcongestion caused immeasurable loss. Study on traffic congestion is mainly for theresearch on the overall running condition of traffic flow. The macro-dynamiccharacteristics of traffic flow refer to the changes of macroscopic parameters (trafficflow and speed). They are easy to obtain and can better characterize the state of thetraffic flow. The macroscopic parameters of traffic flow has strong randomness andnonlinear characteristics, but in the short term also has a certain quasi-periodic.The typical analysis methods about macro-dynamic characteristics of traffic floware divided into a mathematical model of ARIMA and no mathematical model of neuralnetwork. The traditional analytical methods exist many deficiencies. A number ofnonlinear and non-stationary data requires multiple methods. These systems are difficultto express by the exact mathematical model while neural networks do not need toestablish a precise mathematical model, and can achieve good results. In this paper, weuse wavelet neural network to analysis historical traffic flow data. The main researchwork is as follows：①Firstly, introduce the neural network theory and wavelet knowledge, and studythe wavelet neural network, then establish the Wavelet Neural Network Model. Thenpreprocess the collected GPS traffic flow data, including the operations of abnormalrepair data, screening etc. and make simulation experiment for the Wavelet NeuralNetwork Model through data test.②Aiming at the problem of great error, we should improve the wavelet neuralnetwork by adding the momentum and the dynamic learning coefficient at first. Thenanalysis the defects of wavelet neural network, and adpot genetic algorithm to optimizewavelet neural network. As to the crossover and mutation rate of genetic algorithm,adpot dynamic adaptive algorithm and validate the improved algorithm.③Adopt improved statistical algorithms to achieve the matching system of thetraffic flow data and the electronic map. Then use wavelet neural network which optimized by genetic algorithm to simulate the traffic flow macroeconomic parametersdata, and analyze its change laws. Then, introduce the urban road traffic state andanalysis the infulence on the traffic state which made by the variation of macroscopicparameters. Finally, come to the conclusion and future prospect.