期刊文献+

随机森林模型在宏观审慎监管中的应用——基于18个国家数据的实证研究 被引量:14

Application of Random Forest Model in Macro Prudential Regulation-An Empirical Study Based on the Data of 18 Countries
原文传递
导出
摘要 本文以韩国等16个国家的宏观经济数据作为训练集,分别以美国和中国的数据作为测试集,对随机森林这一机器学习模型在系统性风险识别领域的应用进行了探索式研究。结果表明,随机森林模型对系统性风险的识别显著优于传统的逻辑回归模型;对训练集数据学习后的模型不仅通过了稳健性检验,而且具有优异的泛化性能,能够很好地识别中美两国的系统性风险。进一步地,本文引入SHAP机器学习解释模型对风险识别的"黑箱"进行拆解,采用Shapley值和作用力图展示了不同特征变量对两国系统性风险年度概率水平的边际贡献。基于中国的分析表明,稳定信贷杠杆率仍应作为宏观审慎政策的重点;加快金融发展,尤其是深化资本市场,以及稳外资和稳出口对于维护中国金融体系的稳定至关重要。 The identification and prediction of systemic risk is a key step in macro prudential regulation. Since 2017, many financial authorities in developed countries attach great importance to the application of machine learning model in this field. This paper explored the application of the random forest, one of the most influential machine learning model, in systemic risk identification. In this paper, the macroeconomic data of 16 countries such as Republic of Korea is used as the training set, and the data of the United States and China are used as the test set. The results show that the random forest model's performance for systemic risks is significantly better than the traditional logistic regression model;the model not only passes the robustness test, but also has excellent generalization performance and can identify systemic risks in China and the United States very well. This paper introduces the SHAP machine learning interpretation model to dismantle the“black box”of risk identification, and demonstrates the marginal contribution of the different variables for the annual probability level of systemic risk in the two countries in the form of Shapley values. The results show that there are threshold-effects for China's net domestic credit, private sector credit to GDP and domestic savings rate. Once the threshold is reached, the probability of systemic risk in China will be increased sharply. Analysis based on China shows that stabilizing credit leverage should still be the focus of macro-prudential policies, and accelerating financial development, especially deepening capital markets, and stabilizing foreign investment and exports are crucial for maintaining the stability of China's financial system. Besides, moderately positive monetary policy will not increase China's systemic risk probability in the short term.
作者 王达 周映雪 Wang Da;Zhou Yingxue(Institute of American Studies,Jilin University;Institute of National Development and Security Studies,Jilin University;School of Economics,Jilin University)
出处 《国际金融研究》 CSSCI 北大核心 2020年第11期45-54,共10页 Studies of International Finance
基金 2019年国家社会科学基金一般项目“5G通信时代美国对华‘全政府’科技遏制战略与中国数字经济创新发展研究”(19BGJ044)资助。
关键词 宏观审慎 系统性风险 机器学习 随机森林 SHAPLEY值 Macro Prudential Systemic Risk Machine Learning Random Forest Shapley Value
  • 相关文献

同被引文献162

引证文献14

二级引证文献53

相关作者

内容加载中请稍等...

相关机构

内容加载中请稍等...

相关主题

内容加载中请稍等...

浏览历史

内容加载中请稍等...
;
使用帮助 返回顶部