基于支持向量机的贷款风险等级分类真实性审计研究  

Study on the Loan Risk Classification Authenticity Audit Based on Support Vector Machine

作　　者：隋学深[1] 乔鹏[2] 丁保利[3] 

机构地区：[1]审计署审计科研所 [2]审计署金融审计司 [3]审计署卫生药品审计局

出　　处：《审计研究》2014年第3期21-25,74,共6页Auditing Research

摘　　要：商业银行贷款风险等级分类真实性审计,不仅关系到国家审计能否对商业银行信用风险以及系统性金融风险作出正确评估,还关系到国家审计能否有效完成维护国家金融安全的工作任务。构建基于支持向量机的银行贷款风险等级分类真实性审计二分类预测模型,用该模型分析商业银行实际生产数据,得到了86.79%的分类正确率。在验证实验中,100个被故意错分的贷款记录中,分类预测模型识别出了其中的84个。实验结果表明模型在审计实践中具有实用价值。Commercial bank loans risk classification authenticity audit, as for the national audit, can not olny make the correct evaluation of the credit risk of commercial banks and systemic financial risk, but also effectively safe- guard national financial security task. The paper construct two classification predict model of the support vector ma- chine for bank loans risk classification authenticity audit. Based on the actual production data of the commercial bank ,by using the model analysis ,obtained 86. 79% correct classification rate. In the experiments, 100 were inten- tionally misclassified loan records , classification prediction model to identify the 84 of them. The experimental re- sults show that the model has practical value in audit practice.

关 键 词：贷款风险 等级分类 审计 支持向量机 

分 类 号：F239.4[经济管理—会计学]