Scholarlink

<p>Against the background of economic globalization and the goal of “double carbon”, the impact of international energy market fluctuations on the resilience of China’s manufacturing industry is increasingly significant. To clarify the risk linkage between international energy market fluctuations and China’s manufacturing industry, this study takes the international traditional energy (natural gas, crude oil, coal) and clean energy market (wind energy, solar energy, iShares global clean energy ETF) and Manufacturing Purchasing Managers’ index (PMI) as the objects, constructs the “multi-energy manufacturing industry” risk spillover analysis framework, quantifies the time-frequency two-dimensional risk spillover intensity through the Time-Varying Parameter Vector Autoregression-BK (TVP-VAR-BK) model, uses the complex network to visualize the risk transmission path, and introduces the Random Forest (RF) model to construct the risk early warning system. The research findings are as follows: (1) The total risk spillover effect in the time domain is 35.59%. Specifically, clean energy (CLE) contributes a net spillover effect of 12.20%, while the manufacturing industry (PMI) has a net risk reception of -8.41%. Moreover, the short-term spillover (27.19%) is significantly higher than the long-term one (8.41%); (2) During extreme events (such as the COVID-19 pandemic and the Russia-Ukraine conflict), the total system spillover significantly exceeds 40%, and the market linkage exhibits significant time-varying characteristics. (3) The natural gas market (HH) receives risks in the short-term (with a net spillover of -0.25%) and becomes a spillover source in the long-term (with a net spillover of 5.55%), highlighting its crucial role in the energy transition. (4) The Random Forest model performs optimally in early warning (MSE = 0.0966, R² = 0.9627), and the 5-fold cross-validation demonstrates its good robustness. This study has established an interdisciplinary research paradigm of “dynamic measurement analysis-complex network analysis-machine learning early warning”, offering theoretical and methodological support for the manufacturing industry to address energy market risks.</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

Time-frequency Analysis of the Linkage Between the International Energy Market and China’s Manufacturing Industry Risks and Research On Machine Learning Early Warning

  • 期刊论文

  • 发表日期：

    2026年2月3日

查看全文
  • Zijin Cao,
  • Guanghui Han,
  • Hui Xie,
  • Xuedou Yu

摘要

Against the background of economic globalization and the goal of “double carbon”, the impact of international energy market fluctuations on the resilience of China’s manufacturing industry is increasingly significant. To clarify the risk linkage between international energy market fluctuations and China’s manufacturing industry, this study takes the international traditional energy (natural gas, crude oil, coal) and clean energy market (wind energy, solar energy, iShares global clean energy ETF) and Manufacturing Purchasing Managers’ index (PMI) as the objects, constructs the “multi-energy manufacturing industry” risk spillover analysis framework, quantifies the time-frequency two-dimensional risk spillover intensity through the Time-Varying Parameter Vector Autoregression-BK (TVP-VAR-BK) model, uses the complex network to visualize the risk transmission path, and introduces the Random Forest (RF) model to construct the risk early warning system. The research findings are as follows: (1) The total risk spillover effect in the time domain is 35.59%. Specifically, clean energy (CLE) contributes a net spillover effect of 12.20%, while the manufacturing industry (PMI) has a net risk reception of -8.41%. Moreover, the short-term spillover (27.19%) is significantly higher than the long-term one (8.41%); (2) During extreme events (such as the COVID-19 pandemic and the Russia-Ukraine conflict), the total system spillover significantly exceeds 40%, and the market linkage exhibits significant time-varying characteristics. (3) The natural gas market (HH) receives risks in the short-term (with a net spillover of -0.25%) and becomes a spillover source in the long-term (with a net spillover of 5.55%), highlighting its crucial role in the energy transition. (4) The Random Forest model performs optimally in early warning (MSE = 0.0966, R² = 0.9627), and the 5-fold cross-validation demonstrates its good robustness. This study has established an interdisciplinary research paradigm of “dynamic measurement analysis-complex network analysis-machine learning early warning”, offering theoretical and methodological support for the manufacturing industry to address energy market risks.