首页|期刊导航|金融科技时代|集成学习在债务风险预警中的效力

集成学习在债务风险预警中的效力OA

The Efficacy of Ensemble Learning in Debt Risk Early-Warning:Evidence from Chinese Listed Real Estate Companies

中文摘要英文摘要

本文运用2008-2022年89家在A股和港股上市的中国房企数据,检验集成学习模型对上市房企债务风险预警的有效性,研究发现科技驱动的智能监管能有效识别中国房企债务风险.结果表明:本文构建的基于LASSO-GA-XGBoost模型的预警框架,对中国上市房企债务风险的预警表现优于其他模型.股价表现及预售情况是房企债务风险预警指标体系中最重要的维度.合同销售金额、股票年化收益率、股票收盘价和股票年化波动率对房企债务风险预警的重要性均排名前列.本文结论对如何运用人工智能模型完善企业信用风险预警体系,提升金融风险防控的前瞻性,实现智能监管,具有较好的政策启示.

This paper tests the efficacy of ensemble learning in debt risk early warning of listed real estate enterprises.We use 89 real estate enterprises listed in China Mainland and Hong Kong between 2008 to 2022.The results show that the early warning framework based on the LASSO-GA-XGBoost model outperforms others.Stock price performance and pre-sales are the most important indicator dimensions in the debt risk early warning indicator system of real estate enterprises.Contracted sales amount,annualized stock yield,stock closing price and annualized stock volatility all rank high in importance for debt risk early warning of real estate enterprises.The conclusions provide valuable policy insights on how to utilize artificial intelligence models to enhance the enterprise credit risk early warning and monitoring system,improve the foresight of financial risk prevention and control,and achieve intelligent supervision.

钟永红;钟俊豪

华南理工大学经济与金融学院,邮政编码:510006惠州学院数学与统计学院,邮政编码:516007

管理科学

房地产企业债务违约风险预警集成学习

Real EstateDebt DefaultRisk Early-warningEnsemble Learning

《金融科技时代》 2026 (5)

28-42,15

本文感谢国家社会科学基金一般项目(23BJY118)、广东省自然科学基金项目粤惠联合基金青年项目(2024A1515110111)的资助.

评论