【摘要】 党的二十届四中全会强调大力发展“五篇大文章”，着力构建风险防范化解体系以保障金融稳健运行。银行业信贷支持“五篇大文章”已经进入规模跃升阶段，“准确预测相应信用违约风险”及“识别信用违约在五篇大文章中的差异特征”成为该阶段的重要要求，也是落实党的二十届四中全会精神的关键。文章使用机器学习方法识别企业违约风险，进行可解释性分析，并创新地借助大语言模型和因果推断方法识别违约风险与影响因素之间的因果关系，在一定程度上缓解人为设定因果链条所带来的潜在偏误。研究发现：针对科技型企业，投资者情感是违约风险的关键驱动因素；针对绿色型企业，违约风险与地区宏观经济之间存在负向因果关系；影响中小微企业信用违约风险的核心机制在于其现金流是否具有坚实的资产支撑；在养老型企业中，管理层披露的负面情绪是企业违约风险上升的关键信号。实施差异化监管，规范数字金融发展，助推金融供给侧结构性改革，可为金融机构优化资源配置、筑牢重点领域风险防线提供理论依据。更多还原

【Abstract】 The Fourth Plenary Session of the 20 th Central Committee of the Communist Party of China emphasized promoting the “Five Major Articles” and strengthening the risk prevention and mitigation framework to ensure financial stability. As the banking sector’s credit support for the “Five Major Articles” enters a rapid scale-up phase,accurate default-risk forecasting and identification of cross-article heterogeneity have become imperative and are central to implementing the guiding principles of the Fourth Plenary Session of the 20 th CPC Central Committee. This paper assesses corporate default risk using machine learning and provides an interpretable analysis. Furthermore,we innovatively integrate large language models with causal inference to identify the determinants of default risk,mitigating biases from manual causal specification. We find that,for technology-based enterprises,external investor sentiment is the key determinant of default risk. In green enterprises,default risk exhibits a negative causal relationship with regional macroeconomic conditions. For micro,small,and medium-sized enterprises,the core mechanism driving credit default risk lies in whether a firm’s cash flow is supported by a solid asset foundation. In silver economy companies,negative sentiment in managerial disclosures is a key leading indicator of elevated corporate default risk. By implementing differentiated regulation and standardizing digital finance development,financial supply-side structural reform will be advanced. This will offer a theoretical basis for financial institutions in resource optimization and risk management for critical sectors.更多还原