Forecasting the gas prices in Investing.com's weekly economic data table using linear and non-linear ARMA-GARCH models for period 2016-2018

Abstract

There are many time series have large variability, that makes them suffer from the problem of heterogeneity of variance. The analysis of time series requires homoscedasticity in the variance, so for this reason we study some important models that are dealing with a heteroscedasticity time series such as ARMA-APGARCH, ARMA-GARCH and ARMA- NGARCH, which discovered by Engle since 1982. The purpose of the paper is forecast the daily gas prices for the period 2016 to 2018 by using ARMA-GARCH,ARMA-NGARCH&ARMA-APGARCH, when the errors follows a normal distribution and a student T distribution and we use some criteria of selection such as AIC, SIC, H-QIC to select the appropriate model, we conclude that the ARMA (0,1) -NGARCH (1,1) model is the appropriate one the forecast the daily gas prices, by using R program.