Traffic Flow Prediction Based On BP Neural Network

Traffic information prediction is an important foundation for intelligent traffic control,traffic guidance, traffic information services and other ITS subsystems, and also is one of theimportant theory of the intelligent transportation system (ITS) field. Traffic flow predictionhas played a key role in the traffic information prediction. Therefore, the research on trafficflow prediction is premise to develop practical and intelligent traffic forecasting system,which has a very important academic value and practical significance to improve the trafficcongestion problems in China. Over the years, traffic scholars have been to improve thereliability of the predicted traffic information as a research focus.The purpose of the paper is to combine advanced neural network technology with trafficflow prediction closely, and also is to predict traffic flow with neural network. The predictionswere done using the actual data with BP neural network, RBF neural network and waveletneural network through the Matlab platform. On the end, the graphical interface of the BPneural network simulation was designed through the graphical interface developmentenvironment (GUIDE) of Matlab.First, according to Sina survey data, serious traffic problems in our country were showed.On the basis of summing up the measures to solve traffic problems at home and abroad, thepaper focused on the point of traffic flow forecasting. On the basis of research achievementsat home and abroad, the existing traffic flow prediction methods were analyzed, which weredivided into conventional prediction, intelligent prediction and combination forecasting in thispaper. The basic methods of each prediction method were introduced. Secondly, focused onthe artificial neural network theory, we introduced the neural network development,characteristics, structure, and learning theory. And then we analyzed the specific steps of BPand RBF and wavelet neural network. Thirdly, using Matlab to train the actual traffic flowobservational data, the BP neural network, RBF neural network and Wavelet neural networkwere applied to traffic flow forecasting and the predicted results were compared. Finally, thesimulation of the graphical interface of the BP neural network was designed through the Matlab GUIDE development environment.