Research On International Gold Price Forecasting Model Based On Empirical Modal Analysis

In the global economic and financial integration,financial market continues to show a complex phenomenon to explain many of the classic financial theory,the main performance is: the financial market is not ideal market efficient market hypothesis description,high intelligence,strong financial market presents the correlation,tight coupling makes it become a complex the nonlinear dynamic system.The prediction model is built to describe such a complex nonlinear dynamic system,revealing the inherent law of the operation of the financial time series,and showing its evolution mechanism in the world,so that people can prevent financial risks,manage the market and supervise the market.There is no doubt that it is of great practical significance and theoretical value,which is also the significance and background of this study.From NASA Huang e proposed a new signal analysis method,empirical mode decomposition(Empirical Mode Decomposition,EMD).The empirical mode decomposition method does not need to set the base function in advance,and decomposes the time series according to the scale characteristics of the time series.When Fourier transform is applied,the harmonic basis function needs to be set in advance,and the wavelet transform needs to set the wavelet basis function in advance.Empirical mode decomposition(EMD)based on Fu Liye transform and wavelet transform further improves the local characteristics of decomposed time series,which is a more effective data mining preprocessing algorithm.Since empirical mode decomposition method has good characteristics of these processing data,any type of signal can be decomposed by empirical mode method.Therefore,signal processing for nonlinear non-stationary multi-scale characteristic data has obvious advantages.Therefore,the empirical mode decomposition(EMD)method is quickly and effectively applied to many engineering fields after it is proposed.This paper is based on the EMD decomposition method to establish a new combined prediction model: FEPS model(FTS-EMD-PSO-SVR).This model is based on empirical mode decomposition of financial time series(Financial Time Series?Empirical Mode Decomposition,FTS-EMD)?Particle swarm optimization(Particle Swarm Optimization)And support vector regression machine(Support Vector Regression)Model,For nonlinear,non-stationary,multi-scale complex financial time series modeling and prediction,we predict New York futures gold trading price,and empirically study the hot areas in this financial market research.The combination forecasting model proposed in this paper is based on decomposition prediction reconstruction theory,which effectively improves the prediction accuracy and reliability(hit percentage)of the financial time series.This paper selects New York futures gold as an empirical market and data,and establishes seven prediction models of ARIMA,GARCH,BPNN,EMD-BPNN,SVR,EMD-SVR and FEPS,and predicts the short-term trend of closing price.The empirical results show that:(1)EMD-BPNN,EMD-SVR and FEPS models have better prediction results than those of ARIMA,GARCH,BPNN and SVR models.This shows that the nonlinear and non stationary time series of financial forecast method decomposition combination can effectively improve the prediction accuracy,and the introduction of EMD interference method can effectively solve the error of a random sequence,adjacent bands may cause error sequence can not reflect the full time to reflect the sequence information.(2)the new FEPS model proposed in this paper has improved compared with the EMD-BPNN and EMD-SVR models.This indicates that particle swarm optimization algorithm can effectively shorten the training time and improve the prediction accuracy by tracking the best value of the current search to find the global optimum.The prediction model proposed in this paper can provide new methods and references for our country to predict the market price under the current fluctuation characteristics of the international gold price market.