Hybridization Model For Capturing Long Memory And Volatility Of Brent Crude Oil Price Data

The Brent crude oil price indices are typically nonlinear, nonstationary, and non-normal behavior with a long memory and high heteroscedasticity; hence, capturing the controlling properties of their changes is difficult. Subsequently, these phenomena weaken the validity and the accuracy of the result of the forecasting methods. Therefore, this study focuses on the hybridization method to capture long memory behavior and heteroscedasticity in the dataset and improve Brent crude oil price forecasting accuracy. Recently, the hybridization method for the autoregressive fractionally integrated moving average (ARFIMA) model has been introduced as an effective technique for overcoming the nonlinear, nonstationary, and non-normal behavior with high heteroscedasticity in a time series dataset. ARFIMA hybridization method presents several characteristics that other traditional methods do not have. Thus, this thesis proposed three new models and employed 12 different techniques based on combining and hybridizing the ARFIMA model with traditional forecasting techniques to forecast the Brent crude oil price. The three new models, namely, ARFIMA with the asymmetric power autoregressive conditional heteroscedasticity (ARFIMA-APARCH), ARFIMA with the Glosten, Jagannathan, and Runkle generalized autoregressive conditional heteroscedasticity (ARFIMA-GJRGARCH), and ARFIMA with the component standard GARCH (ARFIMA-csGARCH) are proposed. This proposal aims to obtain improved forecasting results and solve the forecasting inaccuracy problem in oil price series