Multi-Scale Noise Reduction Based Wavelet

Abstract

Traffic flow sampled data is noisy and chaotic time series. Complex noise component affects the traffic flow predictability. In this paper，it used multi-scale noise reduction based on wavelet/wavelet packet to shield the traffic flow’s noise components interference in deterministic component.It aimed at the contradiction between similarity and predictability in traffic flow noise reduction process, then proposed multi-state threshold method. Experimental results show that, compare with the traditional threshold value method, the threshold method can more effectively extract the traffic flow’s effective information. This method not only made traffic flow which is reduced noise higher goodness of fit, and the prediction accuracy is higher than the traditional threshold methods, thereby we can significantly enhance prediction performance.