| Crude oil is the world's most important industrial raw material.The development of the world economy is affected by huge fluctuations in crude oil prices to a large extent.Governments and enterprises of various countries are paying more attention to the fluctuation of international oil prices.The supply-demand relationship in the energy market is related to the ever-changing international oil prices,and at the same time is inseparable from geopolitics,financial markets and other factors.This makes the accurate prediction of international oil prices a difficult problem that academia has painstakingly studied to overcome.International crude oil prices exhibit complex nonlinear and non-stationary characteristics such as time-varying autocorrelation,because many factors such as geopolitics,international emergencies and financial speculation all affect the fluctuations of international oil prices.Based on literature research and related theories of energy finance,this paper selects representative crude oil price influencing factors to constitute the data set.Through the analysis of the characteristics of the research object(WTI crude oil spot price),the most suitable 4RNN models are used to predict oil prices under multiple influencing factors.It is found that the LSTM model performs best,and its prediction accuracy exceeds the traditional linear model.Compared with articles using machine learning methods to forecast oil price since 2015,the forecast performance is relatively excellent,which proves that the LSTM model has great potential in crude oil price forecasting.Different from the numerous documents that have used machine learning methods for international oil price forecasts in the past five years,this article is based on multiple influencing factors for oil price forecasting,and discusses which factors can improve the accuracy of oil price forecasting.To this end,a relatively novel method based on machine learning in international crude oil price forecasting is used to evaluate the contribution of various variables to improving the accuracy of oil price forecasting.The results prove that: 1.WTI crude oil futures prices,MSCI global index,S&P 500 index,US crude oil export volume,natural gas spot price,and US dollar index can all improve the accuracy of international crude oil price forecasts.2.Relatively speaking,gold prices,US crude oil production,Dow Jones Industrial Average,and US crude oil commercial inventories have no effect on forecasting accuracy and are of low importance in the context of this article.Finally,through the two large dimensions of data characteristics and model methods,this article is compared and analyzed with related documents,and it is found that no matter from which point of view,the prediction accuracy of this method is in the ranks of relatively excellent.This paper also conducts a detailed comparative analysis of oil price forecasting related literature based on multiple influencing factors,and on this basis,we further discuss the importance of each influencing factor for improving the accuracy of crude oil price forecasting.The research found that: 1.Recognized factors affecting crude oil prices such as MSCI Global Index,Standard & Poor's 500 Index,U.S.crude oil export volume,natural gas spot price and U.S.dollar index will also greatly affect the accuracy of oil price prediction;2.Some documents believe that MSCI is not that relative to oil prices,which may be because the traditional linear model is difficult to extract some nonlinear characteristics of the data;3.Some articles show that the US crude oil inventory is correlated with the WTI crude oil price,but this variable does not improve the prediction accuracy in the context of this article.It may be due to the fact that this article uses both US crude oil export volume and production as explanatory variables.When these supply factors are considered at the same time,the role of US crude oil inventories is likely to be substituted. |
PDF Full Text Request