Research And Application Of ARIMA-LGARCH Model In WTI Index

Crude oil has been called black gold.Crude oil is not only an important basic material and energy in the world,but also an indispensable strategic reserve material.It also restricts the country's economic development and security issues.Therefore,it is very important to predict the price of crude oil to the politics and economy.In order to describe the time series' model more accurately,the paper makes an empirical analysis of WTI index.In addition,the ARCH effect test,ARIMA parameter selection and stability test were carried out for the parameters' selection and stationary test.Further more,the ARIMA-LGARCH model is proposed,which is improved from the ARIMA-GARCH model.The definition of ARIMA-LGARCH model is given,as well as the model is used to predict the WTI index.The ARIMA-LGARCH model can well fit the WTI index and predict the short-term trend by comparing the real value and the predicted value.The results show that the ARIMA-LGARCH model can effectively solve the "tailing" phenomenon and improve the prediction's accuracy.The main contents of this paper are as follows:1.The feature set is built,and the automatic feature selection is realized.The existing methods are mainly concerned with the statistical characteristics of data,such as price,period,price fluctuation,etc.But the derivative characteristics of data were ignored.According to the periodicity of transaction data,this paper analyzed data from the golden bit,Fibonacci,closing price variance etc.The multi-dimension data is extracted from the WTI index data.Then,this article uses Relief algorithm to reduce the dimension,and establish a reasonable feature set.The results show that by combining the Fibonacci bit closing data,variance,golden bit data statistics,time series model can well predict the WTI index.2.According to the characteristics of data and the experience of transaction,concepts of the golden bit and Fibonacci bits of data were introduced.The characteristics of the data were reflected in multi-dimension by extending the feature of transaction data.In addition,the attributes of data were introduced in detail.The feature's set is selected scientifically,also the the Relief algorithm is selected to perform the feature operation according to the running time of the program and the feature contribution rate.3.Under the premise of establishing the feature set scientifically,this paper analyzes the application of ARIMA model and LGARCH model in the prediction of WTI index.We analyze in detail the advantages and disadvantages of the two models and the phenomena and properties in the prediction of WTI index.Finally,this paper uses the compound ARIMA-LGARCH model to avoid the defects of the single model in the prediction of WTI index,and the accuracy is improved greatly.