Combining wavelet transform and Markov model to forecast traffic volume

This paper proposes a novel method to deal with traffic volume time series prediction by combining the wavelet decomposition and Markov model. The process of this approach first decomposes the historical traffic volume into an approximate part associated with low frequency and several detailed parts associated with high frequency by means of the wavelet transform. These new time series are easier to model and predict. Then, a Markov model is modeled to predict each new time series. Finally, the traffic volume is forecasted by summing up all these values. A numerical example on a real traffic volume time series is used to illustrate the effectiveness of this composite model. The test shows that our approach can provide an acceptable prediction value.