Study And Application Of Urban Road Short-term Traffic Flow Dynamic Prediction Method

Real-time, accurate and dynamic short-term traffic flow prediction is the pursuing goal of the intelligent transportation. However, the large amount information of short-term traffic flow and the strong uncertain noise disturbance, coupled with the complex topological structure of urban road network result in that the development of intelligent transportation is restricted by the problem that how to realize the dynamic prediction method of urban road short-term traffic flow. Many prediction methods have been proposed to solve problems. But the real- time and accuracy of the prediction results are not ideal, because most of them do not consider the uncertain short-term traffic flow or the complexity of urban road network interference factors, and dynamic prediction did not meet the actual sense. Short-term traffic flow signal wavelet is decomposed and reconstructed under the Mallat algorithm in this paper which filters out interference noise of short-term traffic flow information and carries the frequency-domain characteristics analysis out, greatly improving the speed of traffic flow information preprocessing and the precision of forecasting result at the same time. O nce more, the support vector regression prediction is combined in the cross section of the city road network more dynamic prediction model in order that urban road short-term traffic flow forecasting becomes more accurate, real-time and effective.Paper aims to research the urban road network in the accurate dynamic short-term traffic flow prediction after the short-term traffic flow informatio n effective signal quickly and efficiently extracted. The main research works of the paper are as follows:1. The relationship and the mathematical model of the basic parameters of three short-term traffic flow, speed and density are analyzed; and the basic characteristics of short-term traffic flow are summarized and the correlation of short-term traffic flow is probed from the aspect of space and time. The predictability analysis method and influence factors of short-term traffic flow are summarized from the aspect of dynamic characteristic.2. Wavelet decomposition and reconstruction is studied for fast, efficient extraction of the key issues about subject information and details of signals for traffic flow due to the strong noise impact traffic signals; The Mallat pyramid multi-resolution algorithm introduced to achieve a rapid decomposition and reconstruction of short-term traffic flow signal; Finally, At last, a detailed solution strategy for the effective extraction of the short-term traffic flow signal is presented.3. The selection of the adaptive parameters, kernel function and embedding dimension of G-P algorithm are discussed in detail with realizing the dynamic forecast of short-term traffic flow as starting point and combining with the characteristics of SVR itself, and proposes the original traffic flow data preprocessing methods so that the short-term traffic flow can achieve the dynamic forecast of high efficiency and accuracy.4. Based on the research of the urban complex network topology, it gives the relevant sections of matrix representation method, and it puts forward the F-AHP fuzzy calibration method to evaluate the influence weight; Finally a dynamic forecasting model of sho rt-term traffic flow in urban road is proposed combining previous research results,and the rationality of the model is verified by experiments.5. The system of Xi ’an for dynamic prediction of short-term traffic flow was designed and implemented under the ThinkPHP framework by using Mallat wavelet decomposition and reconstruction of short-time traffic flow signal, and research conclusion of SVR regression prediction.