PERFORMA PREDIKSI: KOINTEGRASI GARCH-SVR VS. GARCH UNTUK VOLATILITAS HARGA KOMODITAS ENERGI GLOBAL

Prajna Pramita Izati (Department of Statistics, Universitas Diponegoro, Jl. Prof. Sudarto SH, Tembalang, Semarang, Indonesia 50275)
Fariz Budi Arafat (Department of Statistics, Universitas Diponegoro, Jl. Prof. Sudarto SH, Tembalang, Semarang, Indonesia 50275)

Jurnal Gaussian
Vol 14, No 2 (2025): Jurnal Gaussian

Forecasting provides benefits in decision-making, one of which is forecasting the volatility of global energy commodity prices. However, there are challenges in forecasting volatility due to the presence of heteroskedasticity and long-memory effects in the data. Therefore, a combination of the GARCH and SVR methods is needed as a cointegration-based machine learning approach. The aim of this study is to compare the forecasting performance of GARCH and GARCH-SVR for global energy commodity price volatility. The findings indicate that the GARCH-SVR model performs well when volatility data exhibits non-stationary long-memory characteristics, whereas the GARCH model is more suitable when the volatility data shows stationary long-memory characteristics.