Oil Volatility Forecasting and Risk Management: A Hybrid Machine Learning Approach

Volatility forecasting is an essential task for investors, speculators, and market practitioners. In finance, volatility is used in pricing options and financial derivatives, portfolio optimization, and risk management. Among the financial assets and commodities, crude oil ranks high in terms of economic development and investment purposes. However, the crude oil market is very complicated and fluctuates rapidly as many variables can influence the price of crude oil in different ways, such as global demand and supply, geopolitical risk factors, macroeconomic indicators, and the prices of other commodities such as gold and silver. The impact of such variables on crude oil prices could be indirect and non-linear. However, the literature until recently models crude oil volatility using statistical and econometric models which may not provide accurate forecasts due to the complexity of the market. In this research, to forecast crude oil volatility, the data of West Texas Intermediate (WTI) was collected over the period 2000 to 2023 from the Energy Information Administration (EIA) database, and an ensemble machine learning GJR-GARCH-SVR-LSTM model that is capable of recognizing patterns and complicated relationships between the variables was used. The findings of this research indicates that the used ensemble model provides better prediction accuracy than the existing models. Also, unlike the existing literature, another contribution of this study is that the proposed machine learning models were evaluated based on their hedging effectiveness by using the optimal hedge ratio, providing new hedging strategies to investors, crude oil producers, and other market participants.