个人信用评分的Adaptive Lasso-Logistic回归分析  

Analysis of Individual Credit Scoring Based on Logistic Regression with the Adaptive Lasso

作　　者：张婷婷[1] 景英川[1] 

机构地区：[1]太原理工大学数学学院,山西太原030024

出　　处：《数学的实践与认识》2016年第18期92-99,共8页Mathematics in Practice and Theory

基　　金：国家自然科学基金(11571009)

摘　　要：以个人信用风险为研究对象,分析影响个人信用评分的因素.利用某商业银行个人信用数据,并采用.Adaptive Lasso-Logistic回归模型对影响顾客的个人信用风险的因素进行分析,并与传统Logistic回归模型以及Lasso-Logistic回归模型进行比较.以对顾客"好"与"坏"的二分类结果的正确比例为主要衡量标准,实证发现以.Adaptive Lassi-Logistic回归方法建立的个人信用评分模型,在变量选择和解释上,以及预测的准确性上,均优于传统的Logistic和Lasso-Logistic方法.In order to analysis the influencing factors of personal credit scoring,regarding the personal credit risk as the research object.A bank personal credit data was used,and the Adaptive Lasso-Logistic regression model was adopted to analysis the influencing factors of personal credit risk from consumers.What is more,we compare this model with the traditional Logistic model and the Lasso-Logistic model.Regarding the proportion of correct classification results of ＇good＇ or ＇bad＇ consumers as the main measure,the result comparing with the traditional Logistic regression model and Lasso-Logistic regression model,shows that using the adaptive Lasso-Logistic regression model to establish the personal credit scoring model is superior to the traditional Logistic model and Lasso-Logistic model on variable selection and variable explanation and prediction accuracy.

关 键 词：个人信用评分 ADAPTIVE Lasso Lasso-Logistic LOGISTIC模型 

分 类 号：O212.1[理学—概率论与数理统计]