Chinese Crude Oil Futures Price Prediction Based On Machine Learning

Since the reform and opening up,China’s rapid economic development,the demand for crude oil is also increasing.However,due to factors such as resource reserves and mining technology,China’s crude oil production has continued to decline.On the one hand,demand continues to increase,and on the other hand,production capacity continues to decline.As a result,China has become the world’s largest importer and second largest consumer of crude oil.The crude oil industry accounts for a huge proportion of China’s economic system,and crude oil has become one of the important factors affecting national security and economic and social development.China’s crude oil futures were officially listed on March 26,2018.The launch of crude oil futures gave my country’s market a certain degree of oil pricing power.However,because crude oil has the dual attributes of finance and energy,crude oil futures prices will be affected by the international economic situation and the global military environment,and often fluctuate sharply.Crude oil futures prices have complex nonlinear and non-stationary characteristics,and traditional linear time series models have poor forecasting effects.This paper refers to the existing research results and takes the daily trading data of China crude oil futures from March 26,2018 to August 31,2021 as the research sample.Left-biased,the residual series has an ARCH effect.The GARCH family models based on normal distribution,GED distribution and t distribution are respectively constructed.The research shows that the GARCH-GED model has better fitting effect.Finally,this paper uses the information obtained by the GARCH family model as the input variable of the machine learning LSTM model,combines LSTM with the GARCH family model,builds multiple mixed models and predicts the closing price of China’s crude oil futures,and selects the commonly used RMSE,MAPE and MAE three.The error indicator compares the prediction performance of each model.The study found that the prediction effect of the mixed model was better than that of the single model,and the EGARCH-LSTM model in the mixed model had the best prediction effect.According to the analysis results of this article,multiple models are used to predict the price of Chinese crude oil futures.The hybrid model constructed by the GARCH family model and the LSTM model has high prediction accuracy,has a certain degree of effectiveness and practicability,and has control over market investors’ decisionmaking and regulatory authorities.Risks and the formulation of relevant policies by government departments have certain guiding significance.