Energy prices, including those of crude oil, natural gas, and electricity, are inherently volatile due to a wide range of influencing factors, such as geopolitical events, shifts in supply and demand, and fluctuations in weather conditions. These unpredictable movements pose challenges for decision-makers in energy-related industries, including policymakers, traders, and energy companies, all of whom require accurate forecasts to make informed choices. Predicting energy prices with a high degree of accuracy is essential for minimizing financial risks and ensuring stable supply and demand dynamics. This paper investigates the use of advanced statistical and machine learning models to forecast energy price movements more effectively. Specifically, we compare traditional time-series models, such as ARIMA (Autoregressive Integrated Moving Average), GARCH (Generalized Autoregressive Conditional Heteroskedasticity), and VAR (Vector Autoregressive), alongside hybrid models combining machine learning techniques. By integrating time-series characteristics, including seasonality, volatility clustering, and nonlinear behavior, we assess the effectiveness of each model in predicting price movements. The performance of the models is evaluated using standard accuracy metrics, including the Root Mean Squared Error (RMSE) and the Mean Absolute Percentage Error (MAPE), which allow us to compare forecast accuracy. Our findings reveal that hybrid ARIMA-GARCH-LSTM (Long Short-Term Memory) models significantly outperform traditional econometric approaches, excelling in both capturing the mean behavior and the volatility dynamics inherent in energy prices. This paper demonstrates that hybrid models offer superior forecasting capabilities by leveraging the strengths of both statistical and machine learning techniques, thus improving prediction accuracy for energy prices in volatile markets.
Florina Rahman (Wed,) studied this question.