摘要
在互联网金融日益发达的社会背景下,贷款业务在银行流资金链结构中占据的地位不容置疑,贷款业务中个人贷款行为最难预测,且办理的贷款业务以个人贷款最多。因此,一个合理,精确的个人信用评分系统的重要性不言而喻。本文旨在以总金额额度、信用额度、成功借款次数、借款总额、逾期次数、还清笔数、严重逾期次数、月收入、投资类型九个因素进行分析并建立评分模型。由于影响个人信用风险的因素较多,因此选用主成分分析对数据进行降维处理,最后选取了五个累计贡献率达到91%的主成分建立主成分评分模型。将样本在模型中的得分与其真实的逾期次数比较结果中,基本上逾期次数多的人信用为负,准确率为83%。在此基础上,通过Q型聚类分析将样本按照灰色关联度划分为五类,以此作为五个信用等级,并从所分的五类中各自按照灰色关联度在内部选取最低关联度的样本代表信用等级之间的阈值,将四个阈值样本数据代入主成分综合模型中计算得分,以此得分作为评分等级的划分依据。
In the context of increasingly developed internet finance,the role of loan business in the bank's capital flow chain structure is beyond doubt.In the loan business,the behavior of individual loan is the most difficult to predict,and the loan business handled by individuals is the most.Therefore,the importance of a reasonable and accurate personal credit scoring system is self-evident.This paper aims to analyze and establish a scoring model by nine factors including total amount,credit amount,number of successful borrowings,total borrowings,overdue times,number of repaid loans,serious overdue times,monthly income and investment types.As there are many factors affecting personal credit risk,principal component analysis is adopted to conduct dimensionality reduction processing on the data.Finally,five principal components whose cumulative contribution rate reach 91%are selected to establish a principal component scoring model.When comparing the score of the sample in the model with the actual number of overdue times,the credit of the people with more overdue times is basically negative,and the accuracy rate is 83%.On this basis,through Q type cluster analysis the sample according to the grey correlation degree is divided into five categories,as five levels of credit.And from the points in each of the five classes according to the grey correlation degree,the samples with the lowest correlation degree are selected to represent the threshold value between credit grades.The threshold will be four sample data generation in the comprehensive model to calculate scores,which will be used as the basis for grading scores.
作者
李佳成
任姿霖
王攀
Li Jiacheng;Ren Zilin;Wang Pan(School of Economics and Management,Sichuan Technology and Business University,Meishan 620000 China)
出处
《四川工商学院学术新视野》
2022年第1期126-130,135,共6页
Academic New Vision of Sichuan Technology and Business University
关键词
个人信用
主成分分析
灰色关联度
Q型聚类分析
Personal credit
Principal component analysis
Grey relational degree
Q type cluster analysis