Gold Futures Market Research Based On Machine Learning

Futures is an integral part of Chinese financial market.Gold futures is a key component of the futures market.In recent years,with the continuous rise in the spot price of gold,the trading volume of the gold futures market has also increased continuously.Therefore,the influencing factors and predictions on the gold futures price have important significance for both investors and managers.At present,the development of computer technology and statistics has resulted in many excellent algorithms.Among them,machine learning has become popular in the world,and it can be used in all aspects of life.It has also been widely used in financial markets.The article selects a sample size of a total of ten years from November 2009 to October 2019.For impact factors,previous research has mainly focused on the study of basic indicators.In the method of combining the indicator with the basic indicator,21 representative influence factors are selected,and the missing values in the data are filled by means of mean interpolation.In terms of futures price prediction,this article first uses the partial least squares method to make predictions.The average relative error of the prediction results is1.540 %,and the gap between the predicted value and the true value is large.Then,the article uses random forest regression model,and the control variable method is used to find out that when the number of regression trees is 100 and the number of node splits of a single tree is 5,the effect is the best.At the same time,according to the degree of influence of each factor on the gold futures price prediction,a factor importance evaluation was performed,and 7 factors that were important for price prediction were screened out.Finally,the prediction result was obtained.The average relative error between the predicted value and the true value is 0.081 %,greatly improving the prediction accuracy.Finally,the neural network model is used for prediction.Due to some problems such as slow convergence of the BP neural network,the LM-BP neural network model is used and compared with the BP neural network model.The results show that the LM-BP neural network model only needs to iterate 3 times.But the BP neural network needs 400 iterations.The prediction accuracy of the LM-BP neural network model is 0.065 %,which is much smaller than 0.114 % of the BP model.The experimental results show that the LM-BP neural network performs best in predicting the price of gold futures.