Can decomposition of influencing factors improve the ability of models to predict crude oil prices?

Accurate forecasting of crude oil prices is vital for energy companies, investors, and policymakers. Traditional forecasting approaches often concentrate solely on oil price series, overlooking the intricate linkages between underlying influencing factors. This study introduces a novel forecasting framework that applies Variational Mode Decomposition (VMD) not only to the target variable—Shanghai International Energy Exchange (INE) crude oil futures, but also to a broad spectrum of influencing factors. We leverage state-of-the-art machine learning models, construct various independent variable configurations, and examine the effects of secondary screening methods. Our findings indicate that decomposing influencing factors alone does not yield notable improvements in predictive accuracy. However, when the INE series itself is also decomposed and integrated with the decomposed factors, forecast performance improves significantly. In contrast, applying additional screening to the already decomposed factors does not confer further gains. This dual decomposition strategy thus provides a more refined means of capturing complex market signals, ultimately enhancing the predictive power of crude oil futures forecasts. These insights offer valuable guidance for refining production strategies, strengthening risk management, shaping investment and hedging decisions, and informing energy policy in volatile market environments.