基于随机森林算法的ESG评价模型及其应用  

An Efficient Environmental, Social and Governance(ESG) Rating Modeling and Application Using Random Forest Algorithm

作　　者：吴彪 陈倩 WU Biao;CHEN Qian(CITIC-Prudential Asset Management Co.,Ltd.)

机构地区：[1]中信保诚资产管理有限责任公司 [2]中信保诚资产管理有限责任公司风险控制部

出　　处：《金融市场研究》2024年第7期17-28,共12页Financial Market Research

摘　　要：为积极践行我国生态优先、节约集约、绿色低碳的高质量发展道路,探索金融行业的ESG管理实践、做好绿色金融大文章,本文借鉴国际主流ESG评级框架和信用评级领域建模实践,立足我国国情,创新构建了一套基于随机森林算法的ESG评价模型。通过机器学习方法的应用,本文建立起上市企业ESG表现与企业财务绩效以及投资组合收益的逻辑框架,为投资组合构建和风险防控提供了一套新的方法。在比较分析多种深度学习算法后,最终采用了随机森林算法来构建模型,预测上市公司ESG评分。从实证分析结果来看,本文的ESG评价模型具有较好的分层效果和单调性,ESG优选组合能获得优于业绩基准的超额回报,在一定程度上反映了上市公司追求经济价值与社会价值相统一的发展观,较好地契合经济高质量发展要求。Sustainable and green finance has become a major objective for financial institutions that are regularly dealing with the measurement and management of ESG risks. This paper seeks to build quantitative ESG metrics,examining equity investment and risk management in practice. After comparing five deep learning algorithms, we adopt a Random Forest algorithm to build a model and predict the ESG ratings of 4,637 A-share listed companies.Furthermore, this paper establishes a framework for ESG performance and the financial track record of public companies along with ESG-driven portfolios based on China's national conditions. From the results of empirical analysis, the ESG prediction model displays a relatively good discriminating ability and a high degree of accuracy.It can also be noted that the ESG-driven portfolio can obtain excess returns that are better than the benchmark. In addition to contributing to the ESG modeling and portfolio management, the research in this study also supports the compatibility of pursuing economic value and social responsibility at Chinese public firms.

关 键 词：绿色金融 ESG评价模型 随机森林 投资组合管理 

分 类 号：X322[环境科学与工程—环境工程] F832.51[经济管理—金融学] F275