Prediction Of The Price Rise And Fall Of Gold Futures Based On Deep Learning

Gold futures is an investment in the trading market,no matter it is hedging or maintaining financial stability,it plays a very important role,and it is of great practical significance to predict its price volatility.The current research on the futures market,on the one hand,is based on the analysis of market fundamentals,such as the analysis of national policies,commodity supply and demand,speculative psychology,etc.This kind of analysis method is difficult to quantify,has strong subjectivity,and relies on the experience of analysts.On the other hand,it is based on the historical data construction model of the financial market to study the futures market.This method has certain objectivity and strong persuasion,so this paper chooses this method to study the gold futures market.The gold futures price series is a time series,and the financial data generally cluster effect.This paper considers to use AR-GARCH model to carry out the series.Because the gold futures price changes is a very complex nonlinear dynamic system,using traditional econometric model,there are kinds of difficulties,in recent years,machine learning and deep learning in dealing with nonlinear time series has a very good learning performance,in order to facilitate comparison with AR-GARCH model,this paper constructs the single factor based on the historical settlement price of gold futures LSTM model.After data preprocessing,multi-factor SVR model and LSTM model were constructed to analyze their ability to predict the rise and fall of gold futures price.Based on the analysis of this paper,the AR-GARCH model based on the gold futures sequence and the LSTM model based on a single factor are established.The model has a lag and low predictive ability,and there is no significant difference in the predictive ability of the two models to the rise and fall of the future price.However,with the addition of other influencing factors in the financial market,the establishment of multi-factor SVR model and LSTM model can effectively reduce the lag and improve the predictive ability.Finally,the multi-factor LSTM model suitable for the data in this paper is selected to analyze the investment decision and provide reference for investors.