This study aims to investigate the role of blue-economy financial signals and geopolitical risk in improving crude oil price (WTI) forecasting performance.
Using daily data, the authors develop a hybrid forecasting framework based on a conditional Generative Adversarial Network (cGAN), whose generator integrates a Temporal Convolutional Network, Bidirectional LSTM (BiLSTM) and an attention mechanism. The model incorporates exogenous variables capturing dynamics in blue-economy and clean-energy markets, alongside geopolitical risk indicators. Its performance is evaluated against a comprehensive set of statistical, machine learning and deep learning models under both level-based and return-based specifications, using MAE, MAPE, RMSE, R², directional accuracy and Theil’s U.
The results show that the hybrid cGAN model performs well in capturing long-term price dynamics under the level-based specification. However, under the return-based specification, forecasting performance converges across models, and no single approach consistently dominates. Statistical validation using the Diebold–Mariano test indicates that differences in predictive accuracy among advanced models are often not significant. The findings also reveal that sustainability-related variables, including blue-economy ETFs, clean-energy indices, ESG benchmarks and geopolitical risk, contain relevant information for understanding WTI dynamics, reflecting increasing interdependence between energy markets and sustainable finance.
The analysis focuses on daily frequency and one-step-ahead forecasting. Future research could extend the framework to multi-step horizons, incorporate high-frequency data and include additional transition-related indicators such as climate policy uncertainty or carbon pricing measures.
The results provide useful insights for investors, portfolio managers and energy firms by highlighting the relevance of sustainability and geopolitical indicators in forecasting and risk management. Incorporating these signals can support more informed hedging and asset allocation decisions in volatile market environments.
This study provides a novel integration of blue-economy financial assets, sustainability indicators and geopolitical risk within a hybrid generative deep-learning framework for oil price forecasting. It contributes by offering a robust evaluation of model performance under different data specifications and by highlighting the role of environmental and financial signals in commodity markets.
