Can Clean Energy Stocks Predict Crude Oil Markets Using Hybrid and Advanced Machine Learning Models?

Abstract

The volatility of crude oil markets and the pressing need for sustainable energy solutions have sparked significant interest in forecasting methodologies that can better capture market dynamics and incorporate environmentally responsible indicators. In this study, we address the gaps in the literature by proposing novel hybrid approaches based on combining wavelet decomposition with machine learning techniques (ANN-Wavelet and SVR-Wavelet) and advanced machine learning techniques (XGBoost and GBM) with advanced clean energy indicators to predict crude oil prices. These hybrid models significantly advance the field by reducing noise and improving result accuracy. Besides, these approaches were used to determine the best model for predicting crude oil market prices. Additionally, we employed the SHapely Additive exPlanations (SHAP) algorithm to analyze and interpret the models, enhancing transparency and explainability. Subsequently, we applied SHAP to investigate the predictive value of various asset classes, including the volatility index (VIX), precious metal markets (gold and silver), fuel markets (gasoline and natural gas), as well as green and renewable energy indices, about crude oil prices. The results reveal that the wavelet-SVR model demonstrates consistent and robust forecasting performance with low RMSE and MAPE values. Additionally, the GBM model emerges as highly accurate, yielding shallow forecasting errors. Conversely, the wavelet-ANN and XGBoost models exhibit mixed performance, showing effectiveness in the Full Sample but reduced accuracy during the Russia–Ukraine conflict. Notably, green and renewable energy markets, such as CGA and NextEra energy (NEE), emerge as significant predictors in forecasting crude oil prices. This research provides critical guidance amidst the Russia–Ukraine conflict in predicting oil prices by emphasizing the importance of incorporating environmentally responsible indicators into investment portfolios and policy choices.