摘
要
摘要
在“消费升级”“扩大内需”的引导下，超前消费成为一种习以为常的消费
方式，这不仅给商业银行消费信贷的发展注入了强大动力，同时也对商业银行进
行消费信贷风险的管控提出了挑战。大数据和互联网技术的发展为商业银行开展
个人信用风险业务给予了大量数据与技术支持，对于商业银行进行个人风险预警
做了数据铺垫和技术铺垫。如何有效防范商业银行个人消费信贷风险、如何做好
个人信用风险预警、如何对商业银行风险预警不准确、不及时的问题进行解决是
文章的主要研究目的。本文综合了传统个人征信体系和大数据征信体系，结合 ZJ
商业银行的消费信贷风险预警现状构建了 ZJ银行的个人风险预警体系，并建立了
ZJ银行个人信用风险预警模型,给其他的商业银行在解决此类风险预警问题上提
供了一定的参考。
首先，本文梳理总结了个人消费信贷和信用风险预警相关的理论和文献，为
后续研究提供理论与方法。其次是结合传统征信体系和大数据征信体系，采用 AHP
层次分析法，构建了包含三个层次的 ZJ银行信用风险预警指标体系，包括个人特
征因素、信用历史、贷款银行相关因素和风险信息这 4个一级指标，以及年龄、
月收入等 14个二级指标，利用德尔菲法进行指标评估，最后得到指标的权重。并
根据权重值确定“风险预警阀值”，依据可能产生的不同程度的风险警度分别设
置了四个预警区间S＜0.6294、0.6294≤S＜0.8599、0.8599≤S＜0.9521、S≥
0.9521，分别对应为重警、中警、轻警、无警四个信用预警等级。接着，将 AHP
得到的指标权重值和评估值作为 BP神经网络的输入值和输出值，利用神经网络对
模型进行训练及仿真，对比 BP神经网络的预测值与AHP评估值，验证模型的有
效性和准确性。最后对全文进行总结，提出 ZJ银行风险预警的建议措施。
本文研究主要结论有：（1）在大数据征信与传统征信构建的指标体系中，月
收入在指标评分体系中的权重值是最高的，公检法违约违规信息位居第二。客户
关联度和本行业务办理数量这两项指标所占比例排在最后。（2）通过对 ZJ银行
实际客户数据的训练和仿真，证明基于 AHP-BP的风险预警模型能够有效准确地
进行风险预警。
关键词：大数据征信；个人消费信贷；信用风险预警；AHP；BP神经网络
I
Abstract
Abstract
Undertheguidance of"consumptionupgrading"and"expanding domestic
demand", advanced consumption has become a habitual consumption mode, which not
only injectsa strongdriving forceinto thedevelopment of consumercredit of
commercial banks, but also poses a challenge to the control of consumer credit risk of
commercial banks. The development of big data and Internet technology has provided a
lot of data and technical support for commercial banks to carry out personal credit risk
business, and paved the data and technology for commercial banks to carry out personal
risk early warning. How to effectively prevent personal consumption credit risk of
commercial banks, how to do a good job in personal credit risk early warning, and how
to solve the problems of inaccurate and untimely risk early warning of commercial
banks are the main research purposes of this paper. This paper integrates the traditional
personal credit investigation system and the Internet credit investigation system under
the background of big data, combined with the current situation of consumer credit risk
early warning of ZJ commercial bank, constructs the personal risk early warning system
of ZJ bank, and establishes the personal credit risk early warning model of ZJ bank,
which provides a certain reference for other commercial banks to solve this kind of risk
early warning problem.
Firstly, thispaper summarises thetheories and literaturerelated topersonal
consumer credit and credit risk early warning to provide theories and methods for
subsequent research. The second is to combine the traditional credit collection system
and the big data credit collection system, using the AHP hierarchical analysis method, to
construct a credit risk early warning indicator system of ZJ Bank containing three levels,
including four primary indicators of personalcharacteristics factors, credit history,
lending bank-related factors and risk information, as well as 14 secondary indicators
such as age and monthly income, using the Delphi method to evaluate the indicators,
and finally to obtain the weights of the indicators The final weight of the indicators is
obtained. Based on the weights, the "risk warning threshold" is determined, and four
warning intervals are set according to the different levels of risk warning, namely S
<0.6294, 0.6294≤ S < 0.8599, 0.8599≤ S <0.9521, S≥0.9521,which correspond to
heavy warning, medium warning, light warning and no warning respectively. The four
II