Forecasting crude oil futures price with energy uncertainty: Evidence from machine learning methods

Abstract

Energy related uncertainty has significant influence on crude oil market. To explore the influence, this paper investigates the predictive ability of the Energy-Related Uncertainty Index (EUI), over and above standard macroeconomic predictors, in forecasting crude oil prices using an array of machine learning methods. We find that EUI has a significant impact on crude oil prices. Moreover, machine learning methods combined with EUI performed better than the linear regression method due to a lower rate of prediction errors. Among these methods, the Random Forest (RF) model with EUI performs better in the short term, while the Attention-enhanced Long Short-Term Memory (Attention-LSTM) model with EUI has more substantial predictive power in the long term. These empirical results pass a series of robustness tests. Our findings have important implications for both regulators and investors in the crude oil market.