Prediction Of Crude Oil Futures Price Based On EEMD-ARIMA-LSTM Combination Model

Crude oil is one of the chemical energy sources that has a significant impact on the world's economic development,and its importance is beyond doubt.In particular,my country,as a major crude oil importer,is increasingly dependent on overseas crude oil.Due to the dual attributes of financial and energy crude oil,its price often has drastic volatility,which seriously affects the global military environment and the international economic situation.The price of crude oil futures is not only affected by market economic factors,but also affected by various emergencies such as the new crown epidemic,so it has complex nonlinearity and non-stationarity,which makes relevant forecasting problems very difficult.The price of crude oil futures is an extremely important economic indicator in my country's futures market.Effective forecasting of international crude oil futures prices can not only provide effective information for the investment and production decisions of financial investment institutions and small and medium-sized enterprises,but also help formulate national energy strategies and policies.Control the direction of economic development and other aspects to provide reference.International crude oil futures prices are a kind of time series data and are affected by many linear and non-linear factors.Traditional time series forecasting models such as ARIMA model and RNN recurrent neural network model have large errors in the forecasting effect of such non-stationary and non-linear complex time series data.For this kind of non-stationary and non-linear data prediction problem,decomposing the original data,integrating and predicting is an effective method to improve the prediction accuracy,and it has been applied in many industrial fields.This paper adopts the idea of "decompose first,then reconstruct and then combine",decompose the original time series data with the EEMD empirical mode decomposition method,then reconstruct the components based on certain standards,and finally use the ARIMA model and the LSTM model for orderly combination,Respectively predict each eigenmode function and trend item obtained by reconstruction,and establish an EEMD-ARIMA-LSTM combined model to make up for the shortcomings of a single model and improve the prediction accuracy.Taking into account the impact of the new crown epidemic,geopolitics and other factors on the global economy in recent years,this article first applies the more time-sensitive WTI crude oil futures price data from January 2,2009 to October 30,2020,to carry out ARIMA and LSTM single models.Construction and forecasting.Then apply the EEMD empirical mode decomposition method to decompose the original WTI crude oil futures price data to obtain several eigenmode functions and a trend term,and reconstruct the multiple eigenmode functions based on the zero-mean hypothesis test to have specific Significant high-frequency terms,low-frequency terms,the high-frequency terms,low-frequency terms,and trend terms are used as new model data,and based on their statistical characteristics,ARIMA and LSTM models are respectively used for fitting prediction,and the obtained model prediction results are added as WTI crude oil futures price final prediction results,and through the commonly used RMSE,MAE,MAPE three error indicators to compare the prediction results of each model,find the optimal prediction model,that is,the combined model EEMD-ARIMA-LSTM has the best prediction effect.According to the analysis results of this paper,the hybrid model EEMD-ARIMA-LSTM constructed based on the idea of "decompose first,then reconstruct and then combine" has higher accuracy for WTI crude oil futures price prediction than single model prediction and EEMD-LSTM model prediction.To a certain extent,it proves that the combination model established in this article has a certain practicality and effectiveness in the field of crude oil futures price forecasting,and can play a certain guiding significance for financial investors to make decisions,regulatory agencies to control risks,and government agencies to formulate relevant policies.