Scholarlink

<p>This study develops a hybrid framework that integrates Markov switching regimes with Random Forest to forecast sectoral CO<sub>2</sub> emissions in the United States and China during global crises. By capturing regime-dependent volatility and leveraging ensemble learning, the model outperforms traditional econometric and machine learning approaches, particularly in electricity and renewable sectors. Results show that renewable energy demonstrates higher resilience under crisis-induced volatility, while aviation and land transport remain the most vulnerable sectors. These findings provide actionable insights for energy transition and policy design. Future research could extend this framework by incorporating additional predictors such as carbon pricing, mobility indicators, or cross-country spillovers to further enhance forecasting accuracy and policy relevance. By coupling Markov switching regimes with the Random Forest algorithm, this study advances beyond existing hybrid approaches such as ARDL + ML, QARDL + ML, or deep learning models (e.g., LSTM and CNN-RNN). The novelty of our framework lies in explicitly identifying nonlinear structural breaks caused by global crises and then leveraging ensemble learning to predict emissions within each regime. This dual-step approach improves both interpretability and predictive stability compared to pure ML methods, particularly in volatile sectors such as aviation and land transportation.</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

Impact of renewable and fossil energy on sectoral CO2 emissions: empirical evidence from the United States and China

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2025年12月22日

查看全文
  • Riadh Trabelsi

摘要

This study develops a hybrid framework that integrates Markov switching regimes with Random Forest to forecast sectoral CO2 emissions in the United States and China during global crises. By capturing regime-dependent volatility and leveraging ensemble learning, the model outperforms traditional econometric and machine learning approaches, particularly in electricity and renewable sectors. Results show that renewable energy demonstrates higher resilience under crisis-induced volatility, while aviation and land transport remain the most vulnerable sectors. These findings provide actionable insights for energy transition and policy design. Future research could extend this framework by incorporating additional predictors such as carbon pricing, mobility indicators, or cross-country spillovers to further enhance forecasting accuracy and policy relevance. By coupling Markov switching regimes with the Random Forest algorithm, this study advances beyond existing hybrid approaches such as ARDL + ML, QARDL + ML, or deep learning models (e.g., LSTM and CNN-RNN). The novelty of our framework lies in explicitly identifying nonlinear structural breaks caused by global crises and then leveraging ensemble learning to predict emissions within each regime. This dual-step approach improves both interpretability and predictive stability compared to pure ML methods, particularly in volatile sectors such as aviation and land transportation.