Scholarlink

<p>The issue of carbon emissions has become a global focus, and news portals have become important platforms for comprehensively understanding and engaging in carbon reduction events. News portals leverage their information dissemination advantages to transmit information to the carbon market, thereby influencing carbon price fluctuations. Few scholars have conducted feature analysis of carbon reduction news on news portals and considered its role in predicting carbon price fluctuations. Therefore, this paper first crawls 19,946 pieces of carbon reduction news from 2019 to June 2023 and uses machine learning methods to uniquely explore the characteristics of carbon reduction news. Secondly, the SHAP method is used to investigate the impact of carbon reduction news on carbon price fluctuations. Finally, the paper innovatively introduces carbon reduction news indicators into RF, SVR, and LSTM machine learning models to improve the prediction performance of carbon price fluctuations. The research results show: (1) Carbon reduction news can be categorized into six types, covering implementation pathways, policy guidelines, the vision and blueprint of carbon neutrality, green finance, environmental education and awareness, and climate politics, with spatial and temporal distribution differences. (2) Positive sentiment accounts for 92.8% of the overall sentiment in carbon reduction news on news portals, while negative sentiment only accounts for 7%, and this emotional tendency shows a relatively stable trend. (3) SHAP analysis reveals that the volume of news reports is positively correlated with carbon prices, while the emotional index of news is negatively correlated with carbon prices. (4) The carbon price prediction results of the RF, SVR, and LSTM Models show that incorporating carbon reduction news features into the carbon price prediction Models improves the prediction accuracy by an average of 0.13, 0.26, and 0.1, respectively. Among these, LSTM has the highest prediction accuracy, and RF has the best fitting performance. Moreover, the contribution of carbon reduction news on implementation approaches to carbon price prediction reached 0.31, reflecting its important role in carbon price prediction. Therefore, understanding carbon reduction news can provide important references for government policy-making, social publicity, and corporate strategic measures.</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

News-driven carbon trading market: machine learning-based carbon price prediction

  • 期刊论文

  • 发表日期：

    2025年9月19日

查看全文
  • Yu Bai,
  • Qian Zhang,
  • Baohong Zheng

摘要

The issue of carbon emissions has become a global focus, and news portals have become important platforms for comprehensively understanding and engaging in carbon reduction events. News portals leverage their information dissemination advantages to transmit information to the carbon market, thereby influencing carbon price fluctuations. Few scholars have conducted feature analysis of carbon reduction news on news portals and considered its role in predicting carbon price fluctuations. Therefore, this paper first crawls 19,946 pieces of carbon reduction news from 2019 to June 2023 and uses machine learning methods to uniquely explore the characteristics of carbon reduction news. Secondly, the SHAP method is used to investigate the impact of carbon reduction news on carbon price fluctuations. Finally, the paper innovatively introduces carbon reduction news indicators into RF, SVR, and LSTM machine learning models to improve the prediction performance of carbon price fluctuations. The research results show: (1) Carbon reduction news can be categorized into six types, covering implementation pathways, policy guidelines, the vision and blueprint of carbon neutrality, green finance, environmental education and awareness, and climate politics, with spatial and temporal distribution differences. (2) Positive sentiment accounts for 92.8% of the overall sentiment in carbon reduction news on news portals, while negative sentiment only accounts for 7%, and this emotional tendency shows a relatively stable trend. (3) SHAP analysis reveals that the volume of news reports is positively correlated with carbon prices, while the emotional index of news is negatively correlated with carbon prices. (4) The carbon price prediction results of the RF, SVR, and LSTM Models show that incorporating carbon reduction news features into the carbon price prediction Models improves the prediction accuracy by an average of 0.13, 0.26, and 0.1, respectively. Among these, LSTM has the highest prediction accuracy, and RF has the best fitting performance. Moreover, the contribution of carbon reduction news on implementation approaches to carbon price prediction reached 0.31, reflecting its important role in carbon price prediction. Therefore, understanding carbon reduction news can provide important references for government policy-making, social publicity, and corporate strategic measures.