Abstract:
The research aims to forecast time series relying on individual models SVR, ARIMA, and the hybrid model "ARIMA-SVR" through different integration methods applied to global oil price data from January 2004 to December 2023, comprising monthly data with 240 observations and compare its results to identify the best model for forecasting global oil price. The integration methods include the additive hybrid model, the multiplicative hybrid, and the regression hybrid model as hybrid models comparing with single models SVR, and ARIMA models. The results showed that the additive hybrid model, ARIMA-SVR Additive is the best model of all models under studying, as it provides the lowest values of prediction accuracy metrics: MAE, MPE, MAPE, MSE. Using the Ljung-Box test for the resulting series it has the first ranking. the additive hybrid model, ARIMA-SVR Additive as the best model for modeling global oil price data is followed by the regression hybrid model, then the multiplicative hybrid model, SVR, and finally ARIMA