Purpose This study aims to enhance the forecasting of sovereign Credit Default Swap (CDS) spreads in the MENA region by developing a hybrid model that integrates deep learning algorithms with a structural term structure framework. Accurate CDS forecasting is critical for managing sovereign credit risk, especially in politically and economically unstable environments. Design/methodology/approach The paper proposes a hybrid forecasting framework combining Long Short-Term Memory (LSTM) and Gated Recurrent Unit (GRU) deep learning models with the Nelson-Siegel term structure model. The framework is tested on daily CDS data for seven MENA countries (2015–2024), incorporating macro-financial variables such as the S&P/Hawkamah ESG Pan Arab Index, US 10-Year Bond Yield, LBMA Gold Price, and OPEC Basket Price. The models are evaluated using RMSE, MSE, R2, and SHAP interpretability tools. Findings Empirical results show that hybrid models, particularly the GRU–Nelson-Siegel variant, significantly outperform standalone LSTM and GRU models in predictive accuracy, especially during high-volatility episodes in Iraq, Egypt, and Tunisia. The inclusion of macro-financial indicators improves the model's responsiveness to market shocks, enhancing forecasting reliability and robustness under uncertain conditions. Originality/value This research contributes to the financial risk modeling literature by demonstrating the superiority of hybrid deep learning–econometric models in forecasting sovereign credit risk in emerging markets. The study offers a practical and interpretable approach to risk assessment, supporting more informed decision-making for investors, policymakers, and credit analysts operating in volatile geopolitical environments. Furthermore, the practical implications of this study are particularly relevant for policymakers and financial institutions in emerging markets. The hybrid deep learning–Nelson-Siegel framework enhances the early identification of sovereign risk deterioration, supporting proactive debt management and crisis prevention. By providing more accurate and timely forecasts of CDS spreads, the model can help central banks and regulatory authorities implement data-driven policies to strengthen financial stability. Investors can also leverage these predictive insights to optimize portfolio allocation, improve hedging efficiency, and better assess exposure to sovereign credit risk during volatile geopolitical periods.
Drira et al. (Mon,) studied this question.