Font Size: a A A

Research On Crude Oil Price Forecasting And Risk Early Warning Management Based On Multi-Scale Decomposition Of Interval Time Series

Posted on:2021-04-29Degree:MasterType:Thesis
Country:ChinaCandidate:Y Y HuangFull Text:PDF
GTID:2381330629980497Subject:Technical Economics and Management
Abstract/Summary:PDF Full Text Request
As a strategic resource on which human beings depend for survival,oil has an important impact on global political stability and economic security.Almost all production and living activities are related to oil price.Therefore,the dramatic fluctuation of crude oil prices has a great impact on the world economy.In recent years,China's oil dependence on foreign countries has become higher and higher.In 2018,China's oil dependence on foreign countries was 70.9%,far exceeding the internationally recognized warning line of 50%.As can be seen from "BP World Energy O utlook 2019",crude oil will still be an important component of the energy structure,and China's dependence on foreign oil is likely to rise further.Crude oil plays an irreplaceable role in China's economic development.The development of all industries related to production and life is inseparable fro m petroleum resources.Changes in international oil prices will directly affect economic development and production.The trend of international crude oil prices and China's Economic operation is closely related,so studying the fluctuation rule of international oil prices,predicting the future trend of crude oil prices,and conducting early warning management of crude oil prices have certain practical significance for China's economic development.For the study of time series models,the Singular Spectrum Analysis(SSA)method is a relatively new non-parametric data-driven model.This method won't be restricted by traditional research methods.The singular spectrum analysis technology uses time series create a trajectory matrix and use singular value decomposition(SVD)to construct a singular value sequence corresponding to the time series to obtain the corresponding singular value spectrum.Different singular values will reflect the time series of different information,so the original time series will be decomposed into relatively independent and interpretable components using singular spectrum analysis technology.Each component represents different information contained in the separated time series,so SSA technology is also commonly used for preprocessing of traditional time series.When used for noise reduction,the singular value is generally used to determine the noise component.The artificial determination will be dominated by subjective factors to a certain extent.Therefore,there may be overfitting or information loss in the timing after noise reduction.In crude oil price prediction,the noise sequence may contain some useful information,so it is necessary to predict the noise sequence.Finally,the three sub-sequences obtained by the decomposition are predicted by appropriate methods.Based on the SSA algorithm,this paper proposes a new crude oil price prediction framework to improve the prediction accuracy of the time series of crude oil prices and provide technical support for the field of oil price prediction.This article takes WTI weekly crude oil price interval data as the research object,studies the fluctuation rules of international crude oil prices,predicts the future trend of crude oil prices,and conducts a crude oil price risk early warning management study.First,the relevant theories involved in this paper are introduced,and a research model of crude oil price prediction and risk early warning management based on interval time series multi-scale decomposition is proposed.Secondly,the SSA technology is used to decompose the original interval time series.By creating a trajectory matrix and using the SVD method,the original interval crude oil price series is decomposed into trend time series,market volatility time series and noise series.Thirdly,the three sub-sequences obtained by decomposition are tested for stationarity.For stationary noise series,the ARIMA model is used for prediction.For non-stationary interval trend series and market volatility series,the combined prediction method of BPNN & SVM is used to predict and then the final integration forecast result.Then use the five evaluation criteria of MSE,MAE,SSE,MSPE,and MAPE to evaluate the prediction results of the model.The evaluation results show that for the time series of crude oil price ranges,the combined prediction model(SSA-BPNN-SVR-ARIMA)has better prediction effect.Finally,carry out oil price risk early warning management research,according to the five ranges of oil price early warning limits,carry out risk early warning research on the predicted range of crude oil price series,and at the same time put forward relevant risk management countermeasures to the risks that crude oil price fluctuations may bring.The empirical forecasting results in this paper show that using the singular spectrum analysis method to predict the price of crude oil has certain advantages.For the subsequences with different information obtained through the singular spectrum analysis,the prediction based on the characteristics of the subsequences has higher prediction accuracy.It is of theoretical and practical significance to carry out risk warning based on the five oil price warning limits and put forward risk management countermeasures from the two levels of state and enterprise.
Keywords/Search Tags:Interval Oil Price Prediction, Singular Spectrum Analysis, Support Vector Regression, BPNN, Risk Early Warning Management
PDF Full Text Request
Related items