Scholarlink

<p>Escalating global geopolitical conflicts have caused significant fluctuations in the EU carbon futures market, threatening carbon price stability and exposing investors and regulators to substantial financial losses. This study uses two types of geopolitical risks—geopolitical threat risk (GRT) and geopolitical act risk (GRA)—to predict weekly EU carbon futures returns (EUCR). We propose an explainable machine learning framework that combines XGBoost with multi-strategy improved Harris Hawks optimization algorithm (MIHHO) and Shapley additive explanations (SHAP). The results show that our MIHHO-XGBoost model outperforms benchmark models in out-of-sample forecasts. We report the powerful predictive ability of GRT and GRA based on SHAP analysis. The GRA contributes more than GRT and outperforms conventional predictors such as exchange rates and energy prices. By integrating SHAP dependence analysis with locally weighted scatterplot smoothing (LOWESS), we uncover nonlinear marginal effects of GRT and GRA on EUCR and further pinpoint critical thresholds where their impact shifts direction. We also observe that FTSE150 index and natural gas price exhibit the strongest interactions with GRT and GRA, respectively. Both GRT and GRA exert pronounced negative impacts on EUCR during the early phases of the 2022 Russia-Ukraine conflict and the 2023 Israeli-Palestinian conflict. These findings highlight the value of incorporating geopolitical risks into carbon trading decisions and risk supervision frameworks.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Forecasting EU Carbon Market Reactions to Geopolitical Risks: An Explainable ML Perspective

  • 期刊论文

  • 发表日期：

    2026年4月10日

查看全文
  • Yang Cao,
  • Donglan Zha,
  • Ze He,
  • Xiaoying Su,
  • Robert Lee Kong Tiong

摘要

Escalating global geopolitical conflicts have caused significant fluctuations in the EU carbon futures market, threatening carbon price stability and exposing investors and regulators to substantial financial losses. This study uses two types of geopolitical risks—geopolitical threat risk (GRT) and geopolitical act risk (GRA)—to predict weekly EU carbon futures returns (EUCR). We propose an explainable machine learning framework that combines XGBoost with multi-strategy improved Harris Hawks optimization algorithm (MIHHO) and Shapley additive explanations (SHAP). The results show that our MIHHO-XGBoost model outperforms benchmark models in out-of-sample forecasts. We report the powerful predictive ability of GRT and GRA based on SHAP analysis. The GRA contributes more than GRT and outperforms conventional predictors such as exchange rates and energy prices. By integrating SHAP dependence analysis with locally weighted scatterplot smoothing (LOWESS), we uncover nonlinear marginal effects of GRT and GRA on EUCR and further pinpoint critical thresholds where their impact shifts direction. We also observe that FTSE150 index and natural gas price exhibit the strongest interactions with GRT and GRA, respectively. Both GRT and GRA exert pronounced negative impacts on EUCR during the early phases of the 2022 Russia-Ukraine conflict and the 2023 Israeli-Palestinian conflict. These findings highlight the value of incorporating geopolitical risks into carbon trading decisions and risk supervision frameworks.