中文摘要：

伴随股市波动加剧,证券公司融资融券业务信用风险形势日趋严峻。违约概率被用于客户准入管理、风险定价与信贷组合管理等多个方面,处于信用风险管理的基础性地位。本文使用Probit、Logistic与Extreme Value模型,讨论融资融券业务个人客户的违约概率计量。在融资融券业务开展时间较短、数据积累与整理工作尚不完善的情况下,本文重点使用信用账户的资产负债结构信息计量违约概率,并从模型准确性与拟合性的角度,使用ROC曲线与Brier评分方法对三种模型进行比较与验证。研究结果表明：三种模型均具有较强的区分能力,但Extreme Value模型准确性最强,拟合效果最好,最适合于计量融资融券业务的违约概率。

英文摘要：

With the aggravation of the stock market volatility,the margin trading and securities lending business of security firm is facing a stringent challenge and the demand of credit risk measurement. As the fundamental of credit risk measurement,the measurement of probability of default is meaningful for the customer access management,risk pricing and credit portfolio management. The paper studies measurement of default of probability for individual client in margin trading and securities lending business firstly by Probit,Logistic and Extreme Value model. The history of margin trading and securities lending business is short,data accumulation and management is not perfect yet. In this situation,the paper included the information of asset- debt structure of credit account,and tested the accuracy by ROC curve and Brier score,which proved that the high precision of all three models,but Extreme Value model has higher adaptability for measuring of margin trading and security lending business than other models.