Gold Price Analysis And Forecasting Based On Independent Component Analysis

The gold,as a non-renewable resource,includes the properties of commodity,currency and finance.It plays a significant role in the financial and stock markets.Therefore,the influence factor analyzing and forecasting of the gold price becomes a topical issue for academic and industrial circles in 21st Century.However,there are lots of shortcomings in the existing model of the gold price analysis and predic-tion.For example,it exists the phenomenon of modal aliasing and boundary effect in the model of the hybrid from empirical mode decomposition(EMD)and inde-pendent component analysis(ICA).Furthermore,ensemble empirical mode decom-position(EEMD)solved the boundary effect of time series decomposition,but the phenomenon of modal aliasing remains in the model of EEMD-ICA.On the basis of the foregoing,this paper proposes a novel adaptive and non-recursive method to an-alyze and predict the gold price,which incorporates variational mode decomposition(VMD),independent component analysis(ICA)and autoregressive integrated mov-ing average(ARIMA),called VMD-ICA-ARIMA.Firstly,the gold price sequence is transformed into multichannel mixed signals via VMD method.Secondly,we apply ICA method on the mixed signals and separate out the potential signals from the original signals(independent component,IC),and explain its economic significance.Finally,the time series is predicted by ARIMA method.Through the comparative tests,our model outperforms others in terms of analyzing and forecasting the gold price.This paper contains five chapters,the details are as follows:Chapter 1 The basic necessary preparations,including basic theory about VMD,ICA and ARIMA method,are introduced briefly.Chapter 2 The analyzing and forecasting approach,which combines VMD,ICA and ARIMA,is proposed.Chapter 3 The gold price is decomposed into multichannel mixed signals and separated out potential signals.Furthermore,we explains the influence factors of the gold price in the perspective of independent component analysis.Meanwhile,robust regression analysis verified the accuracy of the analysis.Chapter 4 We apply ARIMA model on the time series to predict the perfor-mance of the gold price in next 4 years.The result demonstrates that the gold price will be growing for one year,declining for two years and thens to be recovering.Chapter 5 In terms of iterations number and prediction effect,our model is compared with the benchmark ARIMA model,EMD-ICA-ARIMA model and EEMD-ICA-ARIMA respectively.The result shows that VMD-ICA-ARIMA gets convergence with the least iterations number and achieves better forecasting effect.