摘要
信用风险是目前商业银行面临的风险中最为重要和最为复杂的,新巴塞尔协议要求各国条件的银行通过实施内部评级法来度量并控制信用风险,内部评级法即通过银行收集的客户相历史数据来构建数学模型,测算客户的违约概率进而对客户进行评分。文章针对信用评分模型解释变量维数较高,类型丰富,好坏客户类型数量不均衡等特点,利用广义半参数可加模型对户违约概率进行建模,并将Group LASSO方法应用于模型进行变量选择和估计。实证研究表明本文提出的模型和方法与以往常用的线性logistic回归模型相比,在模型的判别能力和预测能以及解释性和计算效率上均有较大优势。
Credit Risk is the most important and complex risk faced by commercial banks. New Basel Capital Accord require that banks with necessary conditions should implement the Internal RatingBased （IRB） approach to control the credit risk. IRB approach is according to construct a model based on historical data to estimate clients＇ default probability and give them rating. Regarding such features of the credit scoring model as high-dimensional explanatory variable, diverse variable types and the imbalance in the number of different types of customers, this paper discusses using the generalized semi-parametric additive model to model clients＇ default probabilities, and the Group LASSO method will be applied to select and estimate variables. Empirical studies show that compared with the usual linear Logistic model, the model and method proposed in this paper have great advantages in terms of discriminant and prediction accuracy as well as explanatory effect and computational efficiency.
出处
《数理统计与管理》
CSSCI
北大核心
2016年第3期517-524,共8页
Journal of Applied Statistics and Management
基金
国家自然科学基金青年项目(项目号:11301351)
