Modelling and forecasting monthly petroleum crude oil prices using a hybrid time series model

Abstract

Crude oil is a naturally occurring resource composed of hydrocarbons and other organic material. Crude oil price exert a great impact on the global economy. Therefore, modelling and forecasting crude oil prices are essential tasks for government policy makers, investors and even researchers. The objective of this study is to develop a more accurate time series model for the monthly crude oil prices. The data consisted of 241 monthly observations of crude oil prices spanning from April, 1999 to April, 2019. Since the time series of monthly crude oil prices was non-stationary, the first difference data set was used where it proved the stationary by both graphical and theoretical techniques. The best Autoregressive Integrated Moving Average (ARIMA) model was selected by using the criteria of Akaike Information Criterion (AIC), Schwarz Information Criterion and Hannan-Quinn Information Criterion after testing for different ARIMA models. Since Auto Regressive Conditional Heteroscedasticity (ARCH) effect was presented in the crude oil price time series, a suitable model was fitted to capture the volatility clustering. The best model was identified by the lowest AIC values after testing for various ARCH and GARCH (Generalized ARCH) models. Hence ARIMA (1, 1, 0) + GARCH (1, 1) was found to be the best model with lesser root mean squared error of 4.3017. It can be concluded that the combination of ARIMA and GARCH models in handling volatility made hybrid models as the most suitable for analysis and forecasting crude oil prices