摘要
目前我国还没有一套规范的个人信用评分指标体系和方法。本文利用真实的个人消费信贷数据,首先建立了个人信用评分的多元线性判别分析模型和BP神经网络模型,然后将线性判别分析模型的结果与其它变量一起作为输入变量建立了混合两阶段个人信用评分模型。实证研究表明,混合两阶段个人信用评分模型相对于前两种单一模型能同时满足预测精度和稳健性的双重要求,从而,突破了通常应用单一模型于个人信用评分领域的局限。
A standard personal credit scoring index system and method has not been worked out in China now.To begin with,a Multiple Linear Discriminant Analysis Model and a BP Neural Network Model on personal credit scoring,based on real personal consumption credit data,are respectively established in the paper.And then a twophase model is established by applying Neural Network in which the result of the Multiple Linear Discriminant Analysis Model is inputted,along with other variables.The experimental analysis shows that the two-phase model is more stable and more accurate,compared with those two simple models and breaks through the limitation of applying a simple model in personal credit scoring.
出处
《上海金融》
CSSCI
北大核心
2010年第4期90-95,共6页
Shanghai Finance
基金
教育部人文社会科学研究青年基金项目(09XJC790011)资助
关键词
个人信用评分
判别分析
神经网络
两阶段模型
Personal Credit Scoring
Discriminant Analysis
Neural Network
Two-phase Model
