Mixed-Forecast Method Of City Traffic Flow

This thesis studies the prediction method of one street corner short time transportation flow and puts forward a mixed prediction method based on AR model and wavelet neural network. By analysis of the characteristics of the samples from a street corner transportation flow, the conclusion can be drawn that becasue the data of short time transportation flow is non-linear, time-varient and uncertain, using one estimate method cann't meet the requirement of prediction accuracy. Accordingly, this thesis decomposes the transportation flow data in frequency domain. Firstly wavelet tranformation is used to filter transportation flow, and then according to frequency the compositions of the transportation sudden change that results from uncertainty are decomposed to four frequency segments. After this procedure the signal is decomposed into one basic signal serial and four signal serials on differrent frequency, which are all steady signal serials. Then using the AR model and wavelet neural network model on the data of different frequency makes the prediction. By comparison it shows that the method, which replaces the traditional BP neural network model to wavelet neural network model to forecast the decomposed low frequency signal, overcomes the disadvantage of the traditional BP neural network model in the inaccuracy of network construction, low rate arithmetic constringency. And consequently the method improves the precision. The experiment shows that both the model of AR and the neural network can meet the requirements of the high frequent signal, but the calculation of AR model prediction is much less than that of neural network. Therefore, with the similar accurency the model of AR is chosen to precede the estimation of the high frequency data, which can also increases speed of estimation. At last, all the predict results are added up and the high prediction result is obtained. By the simulation of the mixed prediction method and the comparison of coincide degree of the prediction value and the true value, it shows that the mixed prediction method can estimate the single street corner short time transportation flow at high performance. At the end of the thesis, comparing the results of this prediction method to the other prediction methods, it demonstrates the mixed prediction method has the advantage of high predict accuracy and high speed.