Research On The Prediction Of Soybean Meal Futures Price In China Based On Combination Model

China is a super country with a population of more than 1.4 billion,whose agriculture and animal husbandry play a pivotal role.As a by-product of soybean oil extraction from agricultural soybeans,soybean meal is the primary raw material for the production of livestock and poultry feed.It can also be used to manufacture antibiotic raw materials,health food,cosmetics,etc.It is an essential link between China’s agricultural economy and the livestock economy.Moreover,futures with spot guidance value in economic development to enhance market openness and transparency,effective hedging of financial risk function can not be underestimated.China’s soybean meal futures as one of the varieties of the influential end on the global futures market,high market activity,and even turnover was once ranked first in the world.Therefore,an in-depth forecasting study of China’s soybean meal futures price is of great practical importance and can help participants grasp the regularity of movement of the soybean meal futures market in general and make reasonable decisions.Furthermore,it can promote the steady and healthy development of China’s soybean meal futures market,stabilizing China’s agricultural and livestock markets.To some extent,it can also improve the stability of economic market development.At the same time,some studies have shown that the combined model can improve the forecasting accuracy of a single model,so this study applies the combined model to forecast China’s soybean meal futures prices.This paper selects the closing price of the soybean meal futures price index of China’s Dalian Commodity Exchange from January 10,2020,to June 11,2021,as the sample data set and construct a combined model based on ES＿ML to study soybean meal futures prices.Firstly,we apply the ES model to model the soybean meal futures price data,and then construct the residual data between the model fitted values and the actual observed values as the new time-series data,and select the first five days of strongly correlated data are chosen as input variables and the current day’s data are used as output variables.The GBRT and SVR models are applied to model them separately,and finally,the two parts of forecasts are summed to obtain the complete forecast.Comparison of prediction results,it is found that the machine learning model is better than the ES model,the combined model of the ES＿GBRT model is slightly inferior to the GBRT model,and the ES＿SVR model has the best prediction effect,where MAPE=0.47% and RMSE=11.40.The study shows that a suitable combination of models can enhance the forecasting effect in China’s soybean meal futures price forecasting.