A new kind prediction model of traffic flow based on the thought “decomposition-independently predictioncombination”
is described in this paper. First, the traffic flow data are decomposed and reconstructed by using the wavelet
transform technique. And the low frequency outline signal and high frequency detail signal are fitted respectively by using
the three time exponential smoothing method and ARIMA. Then stepwise regression technology is used to establish the
regression equation, in which the primitive traffic flow data are the dependent variable, and high, low-frequency signal’s
fitting data are the independent variables. The confirmation analysis is carried on with traffic flow data from typical roads
in some city urban districts. The results show that the regression equation has a high fitting precision, and is better than the
ARIMA function.