Prediction of Gold Price Based on WT-SVR and EMD-SVR Model

Abstract

The current gold market show a high degree of nonlinearity and uncertainty, in order to predicted the gold price, Empirical Mode Decomposition (EMD) is introduced, the use the EMD orthogonal decompose the special functions into a finite number of independent intrinsic mode functions (IMFs), then Grouping the IMFs according different frequencies, using support vector regression (SVR) to predict each IMF group, at last plus each forecasting value of equal weighted will get the final prediction. Comparative analysis with the traditional practice is relatively mature wavelet transform (WT), WT decompose the function into some signal, then using SVR to predict detail signals and approximation signal, at last plus each forecasting parts will get the final prediction. Empirical studies show that: EMD has more accurate prediction than WT. This method provides a new powerful analytical tool for the gold price Prediction, an important guiding tool in theory and practice.