Short-term traffic flow prediction is one of the systematic key technology of the real-time traffic flow guidance system in ITS. This paper analyses the characteristic of the short-term traffic flow firstly, and then forecasts it properly.This paper explains the important position of short-term traffic flow forecasting in ITS in the first. Direct against the uncertainty of the short-term traffic flow, this paper proposes using Fourier transformation, wavelet analysis, chaotic and fractal analysis in the research of short-term traffic flow. Based on the result of study we concluded that the change of short-term traffic flow is stability and chaotic moving is one of the short-term traffic flow existing form.Based on the characteristic analysis of short-term traffic flow, BP neural forecasting method based on wavelet resolve-reconstructs and RBF neural forecasting method based on phase space reconstruction are proposed. The results of real data computing shows improved methods have good forecasting function.
|