Research On Characteristic Test And Prediction Of WTI Crude Oil Futures Price

Crude oil is one of the important energy and raw materials for modern life and production activities.Its price fluctuation is closely related to the smooth operation of macro economy,the healthy development of capital market and the choice of individual investors.Since the 1980 s,crude oil has been gradually financialized,and the effective operation of crude oil futures market has exerted a profound influence on the production activities of government departments,investors and relevant enterprises.In today’s world,economies increasingly interlinked,the network technology continues to advance,the internationalization of investment,further improve the facilitation of degree and make any changes in international crude oil futures prices will produce far-reaching influence,macro managers hope that crude oil futures market effective operation,achieve the effect of futures price stability,and speculators,I hope to have a chance to realize the excess profit,Therefore,the prediction of crude oil market price and the analysis of the main factors affecting the price fluctuation have become a hot research topic in all circles of society,as well as a difficulty in financial research and industrial field.After entering the 21 st century,under the joint action of various factors such as global economic and political uncertainty events,supply and demand fundamentals,financial crisis and market speculation,international crude oil price changes show more non-stationary and nonlinear complexity characteristics.As one of the most representative prices in the international crude oil market,the trend of WTI crude oil futures price is regarded as the reference of the international energy market.Taking WTI crude oil futures price as the research object,this paper discusses the following questions: Is crude oil futures price predictable? If so,how accurately? What are the main factors affecting WTI crude oil futures price fluctuation? Of all the predictors,which ones are the best predictors? In this paper,the above problems are discussed step by step,and a relatively comprehensive empirical view is provided for the characteristic test and prediction of crude oil futures price.The main research contents are as follows:Chapter 3 describes the dynamic characteristics of WTI crude oil futures price from the perspective of long memory,stability and nonlinearity,and concludes that WTI crude oil futures price is stochastic and unpredictable in some periods and trend and predictable in some periods: Firstly,based on the daily frequency series data of WTI crude oil futures price,the dynamic HURST index is used to diagnose the dynamic long memory characteristics of the daily frequency series of WTI crude oil futures price.The results show that the series is trend and predictable in some time periods,and the crude oil futures market does not meet the efficient market hypothesis.This conclusion verifies that the oil futures market conforms to the adaptive efficient market hypothesis: markets are not always efficient,nor are they always ineffective.Secondly,the stationarity test of daily frequency series data of WTI crude oil futures price is carried out.Stationarity test is the prerequisite of market validity test and time series analysis.In this chapter,the multiscale Lyapunov index in complexity science is introduced into the stationarity test and compared with the traditional ADF test,PP test and KPSS test.The dynamic test results show that the test statistics constructed based on the multi-scale Lyapunov index is a feasible test method for stationarity test,and the daily frequency series of WTI crude oil futures price is a non-stationary time series with most of the time period and a small part of the time period.Finally,the dynamic nonlinear test of daily frequency series of WTI crude oil futures price is carried out based on KS test statistics and BDS test statistics by using alternative data method.The empirical results show that the daily frequency series of WTI crude oil futures price is not random,but contains nonlinear components.This conclusion further supports the view that oil futures market conforms to adaptive efficient market hypothesis and oil futures prices are partially predictable.Chapter 4 takes the images generated by the daily frequency sequence of WTI crude oil futures settlement price as the input feature,and based on the deep belief network,discusses the feasibility of using the images as the classification feature to predict the rise and fall trend of WTI crude oil futures price.The results show that the accuracy of prediction can reach85%.Furthermore,the range estimation of WTI crude oil futures price based on the fluctuation trend is constructed by synthesizing the fluctuation trend prediction of opening price,closing price,highest price,lowest price,trading volume and trading volume.In Chapter 5,based on the sequential and nonlinear characteristics of the daily frequency series of WTI crude oil futures price,the LSTM network model is selected to forecast the price data in the past five working days.The results show that the accuracy of prediction can reach more than95%,indicating that the price in the past reflects most of the market information in time.Furthermore,according to the conclusion of stationarity characteristic test in Chapter 3,the stationary and nonstationary segments of time series were selected to construct LSTM network models for prediction,which verified the non-sensitivity of stationarity to LSTM network modeling for the first time.The results show that as long as the amount of data is large enough,the model can fully learn the characteristics of the training data,and satisfactory prediction results can be obtained no matter whether the data is stable or not.Chapter 6 studies WTI crude oil futures price forecast based on forecast index.For the first time,the prediction index system of WTI crude oil futures price is fully included in the machine learning model.The prediction index system contains 42 influencing factors and technical indicators,including macro supply and demand fundamental indicators,financial investment indicators,emotional indicators and technical indicators.In order to solve the problem of how to integrate different frequency indicators into the same framework,mid AS-Adaboost,a mixedfrequency data model,is constructed in this chapter to test the importance of each indicator to WTI crude oil futures price.The analysis results show that Tech23,which can fully reflect the characteristics of past price fluctuations,has the greatest contribution to WTI crude oil futures price,reaching more than 95%.Considering that the price in the past reflected most of the market information and technical indicators were constructed from the past market prices,in order to eliminate the possible multicollinearity problem between technical indicators and influencing factor indicators,technical indicators are removed in the MIDASAdaboost model framework and only non-technical indicators are retained.The importance of influencing factors to WTI crude oil futures price is tested again.The results show that the yield to maturity of government bonds(LTY),the monthly growth rate of us industrial production index(INPRO)and the us weighted exchange rate(TWI)are the three indexes with the largest weight,which contribute 38.66%,26.16% and 23.79%,respectively.The prediction results based on the prediction indicators show that the fitting degree of technical indicators on WTI crude oil futures price is much higher than that of influencing factors,especially technical indicators Rech23,with the prediction accuracy of more than 95%.Chapter 7 is summary and prospect.Summarize the research results and give the future research prospects.