Application Of Short Term Traffic Flow Prediction Methods

With the continuous development of social economy and traffic transportation, more and more people begin to pay attention to the traffic problems, traffic problems not only affect people's travel safety and efficiency, but also affect the development of society. Intelligent transportation system(ITS) is an effective way to solve the traffic problems in the present stage, especially to solve the traffic pressure and the protection of the environment. Traffic flow is one of the most important research content of ITS, the short-term traffic flow forecasting research is the hot research direction, and get real-time traffic flow data is an effective method for road unimpeded. Therefore, how to obtain accurate traffic flow is the main content of study.Firstly, analysis from the short-term traffic flow theory, mainly research characteristics of short-term traffic flow and its theoretical model, and the advantages and disadvantages of various models was evaluated; In order to change rule of traffic flow time series by using better, so the introduction of the analysis theory, and through the analysis of traffic flow characteristics of self correlation discovery; Self similarity as the entry point to the short-term traffic flow time series, NARX neural network prediction model is established. At the same time, the basic principle of neural network theory and BP neural network are analyzed, through the analysis and comparison of BP neural network and NARX neural network, found in the time series prediction, the superiority of the NARX neural network is higher than that of BP neural network with feedback, and memory function, which is conducive to the time series of historical data in combination with the relationship between the current data.According to the NARX neural network model for short-term traffic prediction is introduced in this paper, the traffic flow data loss as a case to verify its feasibility. NARX neural network, the historical data of traffic flow as the input and output of the training, testing, validation, and then the missing data traffic flow prediction, the predicted value and the actual value of the error analysis, the final verification of the error is small, high precision of prediction results.