Geopolitical risk and energy market tail risk forecasting: An explainable machine learning approach

Abstract

This study develops a forecasting model for energy market tail risk, with a focus on the predictive role of geopolitical risk factors. Using daily energy commodities data spanning from 2000 to 2024, this study evaluates the performance of machine learning models. Results indicate that the Light Gradient Boosting Machine (LGBM) consistently outperforms other models based on key metrics. Robustness tests across different tail risk levels affirm LGBM as the optimal choice for energy market tail risk forecasting. Furthermore, model interpretability reveals that geopolitical risk indicators contribute significantly, with a 19.15 % impact on the forecasting model. Notably, the foreign exchange market, influences predictions by 15 %, while the monetary policy, contributes 12.19 %. Our findings have significant implications for regulators, industry practitioners, and investors seeking optimal tail risk forecasting during geopolitical conflicts.