Research On Commodity Futures Price Forecasting Based On Deep Learning

As the key to investment transactions,price trends are related to the sound development of the market and the vital interests of investors.Therefore,forecasting research on prices has always been the focus of research in the field of financial investment.With the gradual improvement of the position of commodity futures in the financial market,more and more prediction researches on commodity futures prices have been made.And deep learning method as an advanced artificial intelligence technology has been widely used in the financial field,so this paper combines the deep learning method to predict the price of commodity futures,in order to provide some reference for investors' investment and trading.This article first combed the theoretical knowledge of related models in deep learning methods,combined with the features of convolutional neural network(CNN)that can effectively extract features and the characteristics of gated recurrent unit network(GRU)that can process time series data quickly and accurately,CNN-GRU was constructed Combined neural network model,and the combined model and a single CNN,GRU neural network model were used to predict the prices of Shanghai copper,soybean meal and PTA futures varieties.The results show that the prediction accuracy of CNN-GRU combination neural network model is the highest,indicating that the combination model is constructed successfully and has significant advantages.Secondly,in order to improve the accuracy of futures price prediction,this paper intends to add technical analysis indicators and fundamental factor indicators to the input characteristics of the model,but this first requires a study of the effectiveness of the futures market.Therefore,this paper uses the autocorrelation test and GARCH model to conduct an empirical study on the effectiveness of Shanghai copper,soybean meal,and PTA futures markets.The results prove that the prices of Shanghai copper,soybean meal,and PTA futures all have the characteristics of spike and thick tail.The test proved that the price series of the three futures varieties had weak autocorrelation and did not obey the random walk hypothesis.The GARCH model test proved that the price fluctuations of the three futures varieties had long-term memory.In summary,the Shanghai copper,soybean meal and PTA futures markets were Has not yet reached weak effective.Then,on the basis of proving that the Shanghai copper,soybean meal,and PTA futures markets have not reached a weak and effective basis,this article adds technical analysis indicators to the model input variables,and uses the CNN-GRU combined neural network model to evaluate Shanghai copper,soybean meal,and PTA futures Price prediction.After applying PCA to the input variable characteristics,the prediction error is significantly smaller than the prediction error based directly on the pricevolume characteristics.The results show that the technical analysis indicators have a significant effect on improving the accuracy of commodity futures price prediction.Finally,by analyzing the fundamental factors affecting the prices of Shanghai copper,soybean meal and PTA futures,and adding fundamental factors to the model input variables,the futures prices of the three varieties are predicted.The empirical results show that the fundamental factors can be better.Improve the accuracy of futures price prediction.