《基于数据挖掘的信用卡个人客户信用评价研究.docx》由会员分享，可在线阅读，更多相关《基于数据挖掘的信用卡个人客户信用评价研究.docx（49页珍藏版）》请在金锄头文库上搜索。
1、论文题目：基于数据挖掘的信用卡个人客户信用评价研究专业：管理科学与工程硕 士 生：聂雨（签名）：指导老师：苏建军摘要（签名）：自上个世纪 90 年代以来，随着中国经济的发展和内需的扩大，信用卡业务迅速发展起来。信用卡业务的关键在于信用风险的控制，为了更好的防范信贷风险和进一步推动个人消费信贷业务的发展，发卡机构必须建立一套完善有效、科学合理的个人信用评价体系。信用评价将客户分为“好”客户和“差”客户，根据历史上每个属性的若干样本，建立数学模型，预测信用卡使用者的违约风险。决策树算法简单直观，误差率低，本文选取该算法作为建立信用评价模型的方法。本文首先对数据挖掘、信用卡业务、信用评价三者的理论框
2、架进行结合，分析了数据挖掘、信用卡业务中信用评价的特点，为下文建模方法的筛选奠定理论基础。其次结合数据挖掘几种方法的不同特性，选择决策树 C5.0 算法作为本文模型建立方法，并详细介绍了如何利用 C5.0 算法建立模型。最后从某发卡行获取的商业银行信用卡个人数据出发，采用上述算法，经过商业理解、数据理解、数据准备、建立模型、模型评估与分析等步骤，建立了个人信用评价的决策树模型。并且依据发卡机构的实际需求，对决策树模型的成本矩阵和修剪程度进行了调整，形成了修正改进的模型。基于决策树方法的个人信用评价模型精确度高、可控性强，可在实际中广泛运用。且通过建立误判矩阵，使得发卡机构运营中成本最低。在实际
3、运用中对于进行信用卡申请者的判别有一定的指导性作用，并能够为信贷决策提供支持，具有较强的理论和现实意义。关 键 词：数据挖掘；决策树；C5.0；信用卡研究类型：应用研究SubjectSpecialty: Research on Credit Rating of Consumer client of CreditCard based on Data Mining: Management Science and EngineeringName: Nie Yu(Signature)Instructor : Su Jianjun(Signature)ABSTRACTSince. the 90s, wi
4、th the development of Chinas economy and the expansion of domestic demand,Credit Card Industry developed rapidly. The key of Credit Card Industry is control of credit risks. Inorder to guard against credit risks efficiently, to promote the development of the consumer creditbusiness, card issuers mus
5、t establish a set of effective scientific and reasonable individual credit system.The credit rating system divided customers into “good” customers and ”bad” customers. Then accordingto the historical data of different attributes, set a mathematical model, forecast the default risk of creditcard user
6、s. Decision Trees is simple, easily understanding and has low error rate, so this paper take thealgorithm as the modeling method.Firstly, this Paper combines data mining, credit card industry, credit rating together, analyzed thecharacteristics of data mining and credit rating for credit card, lay a
7、 theoretical foundation for thefollowing discussion. Secondly, consider the different feature of means of data mining, choose thealgorithm C5.0 of Decision Tree as the following data mining mean, then introduce the specificprocesses of modeling with C5.0. Finally, make use of the individual credit c
8、ard data which comes fromcertain bank, go through business Understanding, data understanding, data preparation, modeling,modeling evaluation, set a model based on decision tree. Then according to the specified actual demandof card issuer, adjust the Cost-sensitive Tree and prune decision tree, form
9、a advanced model.Credit rating of consumer client of credit card based on decision tree has higher accuracy andstronger controllable nature, which can be used widely, and attributable to the cost-sensitive tree, the costof card issuer is minimize. It has directive function in practice for distinguis
10、hing credit card applicant, Ithas strong theoretic and practical significance to provide support for credit decisions,Key Words: Data Mining; Decision Tree; C5.0; Credit CardThesis: Applied Study目录目录1 绪论 .11.1 选题背景与研究意义.11.1.1 选题背景 .11.1.2 研究意义 .31.2 国内外研究现状 .41.2.1 国外研究现状.41.2.2 国内研究现状.61.3 论文框架 .72 数据挖掘、信用评价、信用卡业务综述.92.1 数据挖掘概述 .92.1.1 数据挖掘概念.92.1.2 数据挖掘的体系结构 .92.1.3 数据挖掘功能. 112.2 信用评价概述 .122.2.1 信用评价定义与基础 .122.2.2 信用评价方法综述 .
