Research On Forecasting And Comparison Of Gold Futures Price Based On VMD-wavelet Denoising

Gold has long been in the spotlight for its ability to pay,circulate,store,value scale,and world currency.Because of the complexity of its properties,gold prices are influenced not only by supply and demand,global economic conditions,and national policies,but also by the dollar index,the dollar index,the Dow Jones,and crude oil prices.Therefore,in-depth study of gold prices can help us to better predict gold prices,not only to help countries adjust economic policies in a timely manner but also to help individuals and institutions to control financial risks,has certain theoretical and applied value.The time series of gold futures is a complex system with nonlinear relationships.With the continuous development of the futures market,its development trend is becoming more and more complex.The traditional forecasting model gradually becomes only a certain theoretical value,but it is difficult to It has been effectively used in actual market analysis.This thesis takes the closing price of the main Shanghai gold contract(AU0.SHF)on each trading day as the research object.Based on the idea of multidisciplinary fusion,this thesis introduces variational mode decomposition(VMD)and wavelet denoising methods for signal processing to reduce the influence of noise in the data on the prediction results.First,the original gold futures price time series is decomposed by VMD,and then the correlation analysis between the decomposed component sequences and the original sequence is carried out to find out the high-frequency noisy components,and then the high-frequency noisy components are reconstructed.The wavelet threshold denoising process is performed on the constructed sequence,and finally all components are reconstructed and substituted into the SVR model and the LSTM model respectively to construct a VMD combination model based on wavelet denoising.At the same time,in order to reflect the combined advantages of VMD and wavelet denoising,this thesis also constructs the EEMD-SVR model of wavelet denoising and the EEMD-LSTM model of wavelet denoising for comparison.Finally,the empirical research results show that:(1)The prediction effect of the nonlinear model is better than that of the linear model.(2)The prediction effects of the two combined models are better than the original single model.(3)Among the combination models,the prediction effect of VMD combination model based on wavelet denoising is improved compared with that of EEMD combination model based on wavelet denoising.Among them,VMD-LSTM model based on wavelet denoising has the most accurate prediction of gold futures price.