The Urban Traffic Flow Prediction And Application

With the acceleration of city urbanization and the rapid development of economy, theurban population, motor vehicle quantities are continued rapidly increasing. The trafficproblem increasingly escalated, and it has become a major issue which constrained thedevelopment of the city. Intelligent Transportation System (ITS) arises which become effectiveto resolve the traffic problems. Traffic flow data is indispensable information of ITS.Meanwhile accurate prediction of traffic flow is an important part of ITS, and is also the basisand premise of effective traffic flow guidance and traffic incident detection. So the traffic flowpredictive has important theoretical significance and practical value.Traffic flow is multi-variables, time varying, and complicated nonlinear system. Thesingle traditional forecasting model can only contain part characteristic of the system, soprediction accuracy is limited. In view of this, this article puts forward the combinationforecast method of urban traffic flow based on related intersection analysis and wavelet neuralnetwork. The article deeply researched the prediction model, prediction method and theimplement. The main work has the following contents:(1) This article researches the pretreatment of original data, and puts forward thejudgment method and treatment process of wrong data, loss data and redundancy data. So thenoise data interference can be reduced to a minimum. The pretreatment ensure the datacontains as much as possible of traffic flow information, so as to improve the efficiency andaccuracy of the prediction.(2) The article establishes the neural network forecast model. Then uses the wavelettransform to improve and optimizes neural network parameters, so as to improve theprediction precision of the model and real-time.(3) This paper puts forward traffic correlation analysis method——principal componentanalysis(PCA). And use historical data to analysis the correlation between each intersection.Especially when some intersection’s data was lose, the model still can work.(4) Then this paper put forward the combination forecast model based on neural network,considering the advantages of single point prediction model and the prediction model based onrelated crossings.Paper establishes area intersection network of Guangzhou Tianhe, considering trafficflow quantity and crossroads interval length, etc. Then analysis and preprocess the data of intersections. Finally this article designs combination prediction model based on waveletBP-ANN (Back Propagation neural network) by simulation software MATLAB. In this way,predict the target intersection traffic flow in5minutes by itself historical data and relativecrossroads historical data respectively. This article verifies the combination prediction methodis effective and feasible by simulations.