Study Of The Forecasting Model Of Short-term Traffic Flow Based On Wavelet Transform

With the accelerated urbanization and a large increase in vehicle, some problems, for example traffic jam, personal injury, environmental pollution, are becoming the significant bottlenecks for social economy and culture development, which arise many countries'and regions'concern. Currently, ITS has been developing rapidly as a valid solution to traffic problems. What's more, the traffic flow prediction is the foundation and key factor of the vehicle control and navigation, which can solve the series of traffic problems. In the paper, the forecasting model of short-term traffic flow based on wavelet is focused on.The paper presents main research contents as follows:1. Preprocess the traffic flow data which the detectors collect. According to the characteristic of the traffic flow, scan the time series and identify the missing or error data by maximum threshold method and average vehicle effective length method. Then repair the abnormal data with the linear interpolation method, which ensures the completeness.2. Improve the traditional forecasting model of short-term traffic flow based on wavelet. Firstly, analyse the predicting principle of the tradition short-term traffic flow forecasting model based on wavelet, and illustrate the shortage of it. Secondly, smoothing the traffic flow data with ES in order to get rid of the high noise and increase accuracy of the data.3. According to the principle of the wavelet threshold denoise, improve the soft threshold denoise method. Structure high-order threshold function whenω<|δ|, so that theη(ω) curve becomes smoothed and continuous. Thus, the connected section of the noise coefficients and the valid ones appears more natural, and in accordance with the continuation property of traffic flow.4. With the traffic flow data of Dongzhimen Qiaobei in Beijing, take a demonstration study on the improved forecasting model and the wavelet threshold denoising method. Comparing the error criterions of the traditional model and the improved one, the data proves that the improved model can predict the short-term traffic flow more precisely.Otherwise,the threshold denoising is good for forecasting traffic flow.