Abstract

The current gold market shows a high degree of nonlinearity and uncertainty. In order to predict the gold price, Empirical Mode Decomposition (EMD) was introduced into Support vector machine (SVM). Firstly, we used the EMD method to decompose the original gold price series into a finite number of independent intrinsic mode functions (IMFs), and then grouped the IMFs according to different frequencies. Secondly, SVM was used to predict each IMF in which particle swarm optimization (PSO) was applied to optimize the parameters of SVM. Finally, the sum of each IMF's forecasting result will be the final prediction. In order to validate the accuracy of the proposed combination model, the London spot gold price and the Shanghai Futures gold price series were employed. Empirical studies indicated that the EMD-PSO-SVM model outperformed the WT-PSO-SVM model, and was feasible and effective in gold price prediction. We can promote the EMD-PSO-SVM to other related financial areas.