一种新的银行信用风险识别方法:SVM-KNN组合模型  

A New Method for Banks’ Credit Risk Identification: SVM-KNN Combination Model

作　　者：申晴 张连增 SHEN Qing;ZHANG Lianzeng

机构地区：[1]南开大学金融学院

出　　处：《金融监管研究》2020年第7期23-37,共15页Financial Regulation Research

基　　金：教育部人文社会科学重点研究基地重大项目“基于大数据的精算统计模型与风险管理究”(项目号:16JJD910001)的资助。

摘　　要：科学、高效的信用风险模型是监管机构和银行进行风险管理的重要工具。本文根据SVM模型和KNN模型在处理分类问题中的优势以及二者之间的联系,提出了SVM-KNN组合模型,并以我国165家上市企业2017—2018年度的财务数据为样本,对我国商业银行信用风险进行了识别分析。实证结果表明:SVM-KNN组合模型在识别样本企业信用风险时,不仅可有效解决SVM模型和KNN模型的维度难题,而且相对于单独使用SVM模型和KNN模型,对商业银行信用风险的识别效果也得到了明显提升。最后,本文提出了加强授信准入管理、完善和健全我国银行征信系统和企业数据库系统、实现金融与科技的深度融合等相关建议。Scientific and efficient credit risk model is an important tool for regulators and banks to carry out risk management.Based on the advantages of SVM and KNN model in dealing with classification problems and the relationship between them,this paper puts forward the SVM-KNN combination model.Taking the financial data of 165 enterprises in China from 2017 to 2018 as a sample,this paper uses the combination model to identify and analyze the credit risk of commercial banks in China.The empirical results show that:SVMKNN combination model not only effectively solves the dimension problem of SVM and KNN model,but also improve the recognition effect of commercial bank credit risk compared with SVM model and KNN model alone.Finally,this paper puts forward some suggestions,such as strengthening credit access management,improving and perfecting the credit information system of China's banks and enterprise database system,and realizing the deep integration of finance and technology.

关 键 词：商业银行 信用风险 SVM-KNN组合模型 

分 类 号：F832[经济管理—金融学]