MBA学位论文
作者：李轩
中国银联甘肃分公司商户欺诈风险评级体系研究
中国银联甘肃分公司商户欺诈风险评级体系研究
中文摘要
银行卡收单是一项高收益伴随着高风险的业务。随着我国支付清算市场的快
速发展，特别是非金融支付机构的大量涌入，使得银行卡收单业务的市场规模呈
现逐年上升的趋势。在提供了方便、快捷支付体验的同时，银行卡收单业务面临
的各类风险也逐渐显现，尤其是近年来频发的信用卡套现、银行卡信息泄露等风
险事件，扰乱了金融市场的稳定，对人们的正常生活产生了较大影响，也损害了
银行卡收单业务各参与方的利益，给社会带来了严重的金融风险隐患。中国银联
甘肃分公司于 2008年成立，虽然在公司日常经营过程中已经将商户欺诈风险管
理作为风险管理的主要工作内容，但是由于商户欺诈风险事件为多风险因素共同
导致，公司经营过程中没有有效的、成体系的风险评级方法，因故商户欺诈风险
的管理存在识别效率低、准确度低，管理针对性不强等问题，管理过程缺乏科学
性，没有具体量化的指标作为参考。
本论文以中国银联甘肃分公司辖内收单商户欺诈风险管理为研究内容，通过
利用客观、科学的方法，构建中国银联甘肃分公司商户欺诈风险评级体系。本论
文基于金融风险管理理论、全面风险管理理论，分析了当前商户欺诈风险评级方
法的现状和不足；通过整理中国银联甘肃分公司历史商户数据，梳理出目前可以
收集到的可能导致商户欺诈风险的风险变量，利用 Logistic回归的方法在全部
收集到的风险变量中筛选出影响较大的 28个关键风险变量，并赋予相应的权重，
以此构建商户欺诈风险评级体系；通过设置验证组，在验证组中选取样本进行效
果验证，结合中国银联甘肃分公司的经营现状和商户欺诈风险管理实际，制定出
该欺诈风险评级体系的实施方案和保障措施。
论文研究表明，通过构建评级体系，相较于之前仅依靠主观判断的方法有着
明显的优势，且操作过程更简单、高效，识别出欺诈风险事件的准确度更高、针
对性更强，有助于提升中国银联甘肃分公司商户欺诈风险管理的量化水平，在商
户欺诈风险管理方面提供了方法参考，并且对进一步防范系统性金融风险，提升
金融稳定性有重要的参考意义。
关键词：银行卡收单商户，欺诈风险，Logistic回归，评级体系
I
MBA学位论文
作者：李轩
中国银联甘肃分公司商户欺诈风险评级体系研究
Research on Merchant Fraud Risk Rating System
of China UnionPay Gansu Branch
Abstract
The bank card acquiring business is a business with high returns and high risks.
With the rapid development of my country's payment and clearing market, especially
the influx of non-financial payment institutions, the market size of bank card acquiring
business has shown an upward trend year by year. While providing a convenient and
fast payment experience, various risks faced by the bank card acquiring business have
gradually emerged, especially in recent years, the frequent occurrence of credit card
cash out, bank card information leakage and other risk events have disrupted the
stability of the financial market. It has a great impact on people's normal life, also
damages the interests of all parties involved in the bank card acquiring business, and
brings serious financial risks to the society. China UnionPay Gansu Branch was
established in 2008. Although merchant fraud risk management has been taken as the
main work content of risk management in the company's daily operation process,
because merchant fraud risk events are caused by multiple risk factors, the company's
operation process is not effective. There are some problems such as low identification
efficiency, low accuracy, and poor management pertinence in the management of
merchant fraud risk for some reason. The management process is not scientific, and
there is no specific quantitative index for reference.
This dissertation takes the fraud risk management of acquiring merchants within
the jurisdictionof China UnionPay GansuBranch as the researchcontent, and
constructs the merchant fraud risk rating system of China UnionPay Gansu Branch by
using objective and scientific methods. Based on the financial risk management theory
and comprehensive risk management theory, this paper analyzes the current status and
deficiencies of the current merchant fraud risk rating methods; Risk variables, use the
method of Logistic regression to screen out 28 key risk variables with greater impact
from all the collected risk variables, and assign corresponding weights to build a
merchant fraud risk rating system; Samples were selected from the group for effect
verification, and the implementation plan and safeguard measures of the fraud risk
II
MBA学位论文
作者：李轩
中国银联甘肃分公司商户欺诈风险评级体系研究
rating system were formulated in combination with the operating status of China
UnionPay Gansu Branch and the actual fraud risk management of merchants.
The research of the thesis shows that by constructing a rating system, it has
obvious advantages compared with the previous method that only relies on subjective
judgment, and the operation process is simpler and more efficient, and the identification
of fraud risk events is more accurate and more targeted, which is helpful for In order to
improve the quantitative level of merchant fraud risk management of China UnionPay
Gansu Branch, it provides a method reference for merchant fraud risk management, and
has important reference significance for further preventing systemic financial risks and
improving financial stability.
Keywords: Bank card acquiring merchant, Fraud risk, Logistic regression, Rating
System
III