Research On Agricultural Futures Price Prediction Based On LSTM Neural Network Combination Model

Agricultural products futures trading market is an important part of China’s market economic system,which is of great significance in price discovery,hedging and risk avoidance.Apple futures market has become the most important sub-market in China’s agricultural products futures market.As the first fruit futures variety launched in China,it plays a certain reference significance for China to explore and open other categories of fruit futures.It has been more than five years since apple futures varieties were introduced into China’s agricultural futures market.Because of its futures attributes of price discovery and hedging,the field of agricultural economy is currently studying apple futures price changes and predicting the price trend of apple futures.However,the apple futures market fluctuates frequently due to the multiple influences of spot price,market supply and demand imbalance,seasonal change,information asymmetry and weather and climate factors,which brings new challenges to China’s apple industry and investors.The effective prediction of the trading price of the apple futures market can help Chinese apple enterprises and farmers to avoid the risk of apple market price fluctuation,such as adjusting the scale of production and operation in time,improving the planting operation structure as well as giving play to the hedging role of the futures market,and finally to ensure the stable development of the apple market industry in China.As the time series data,the market price of Apple futures is characterized by non-stationary and non-linearity.The data indicators selected in this paper mainly include the basic price index of apple futures,technical indicators and macroeconomic indicators.The data used in the basic price indicators include the index data of the main contract of Apple futures in the historical trading data of Zhengzhou Futures Exchange from the opening of December 22,2017 to December 31,2021.The macroeconomic data are from the CSMR Chinese database.On the basis of the Adam gradient descent optimization algorithm theory,the apple futures price data is predicted by using the long-term and short-term memory neural network model(LSTM model)and the principal component analysis dimension reduction algorithm.Firstly,I reduce the dimension of the main component data and the closing price of the apple futures.Secondly,I fit and predict the closing price of the apple futures.Thirdly,I output the model fitting results,and the prediction fitting effect of the model is measured by R~2and RMSE(Root Mean Square Error).Finally,comparative analysis with other models.Comparing PCA-LSTM combination model and long-term memory neural network model(LSTM),Linear Regression(LR),Random Forest(RF),Light GBM and XGBOOST,using R~2and RMSE as evaluation indicators.It is concluded that PCA-LSTM has the best prediction effect on the main contract of Apple futures,and has a great advantage in apple futures price prediction.