Abstract:Based on the advantage of the wavelet analysis for non-stationary time series forecasting, this thesis presented multiresolution analysis theory as well as detecting singularity for time series, then made static and dynamic estimation of the processed high and low frequency data by using the autoregressive moving average(ARMA) model and Holt-Winters-No Seasonal with the software of AMTLAB and EVIEWS. The results show that the wavelet analysis in non-stationary time series prediction has a great advantage.

Key words:iron ore,freight rates forecast,wavelet analysis,autoregressive moving average(ARMA) model