基于数据挖掘的商业银行贷款信用评级

发布时间：2019-03-09 17:48

【摘要】：风险管理是商业银行经营过程中的核心,是商业银行健康发展的重中之重。存贷款业务是商业银行经营过程中最重要的组成部分之一,贷款业务是商业银行盈利的主要手段,因此,贷款风险成为商业银行的最主要风险,需要积极合理的应对。本文使用数据挖掘中决策树分类技术,对已获得的某商业银行真实贷款信息样本进行纵向的挖掘。经过数据的获取、集成、预处理,然后建立贷款风险的决策树评级模型,然后以此为基础,得出贷款的风险评级,帮助商业银行风险管理部门预测贷款的风险。由目前商业银行获得的贷款企业的信息可以看出,相关数据大而繁杂,不同企业的各项指标都是无关的,而且完全的相同的指标几乎没有。因此针对传统决策树算法对于处理连续属性的一些局限性,重点研究了应用于商业银行贷款风险预测系统的财务数据离散化方法以及决策树生成方法：对于连续属性,采用聚类的方法对其属性中数据进行基本的划分,然后利用模糊集方法解决连续数据的离散化中的苛刻阈值不公平问题；针对产生的具有隶属度的新属性,使用改进C4.5决策树分类方法,并将其在银行风险预测中应用。使用基于某商业银行系统中的贷款企业详细数据作为训练数据源,结合数据挖掘的思想及方法,应用改进的决策树分类技术,完成在商业银行贷款的企业的贷款风险预测。结果表明,使用改进的方法能够在保证预测评级准确性的同时,使决策树更加简洁,结果更有借鉴意义。
[Abstract]:Risk management is the core of the business process of commercial banks and the most important of the healthy development of commercial banks. Deposit and loan business is one of the most important components in the business process of commercial banks. Loan business is the main means of profit of commercial banks. Therefore, the loan risk becomes the most important risk of commercial banks, which needs to be dealt with positively and reasonably. In this paper, the decision tree classification technology in data mining is used to mine the real loan information samples of a commercial bank. After data acquisition, integration, pre-processing, and then the establishment of a decision tree rating model of loan risk, then based on this, the risk rating of the loan can be obtained, which helps the risk management department of commercial banks to predict the risk of the loan. From the information of loan enterprises obtained by commercial banks at present, it can be seen that the relevant data are large and complicated, all the indicators of different enterprises are irrelevant, and there are almost no identical indicators. In view of the limitations of traditional decision tree algorithm in dealing with continuous attributes, this paper focuses on the discretization method of financial data and the decision tree generation method applied to commercial bank loan risk prediction system: for continuous attributes, The clustering method is used to divide the data in its attributes, and then the fuzzy set method is used to solve the problem of harsh threshold unfairness in the discretization of continuous data. According to the new attribute with membership degree, the improved C4.5 decision tree classification method is used, and it is applied to the bank risk prediction. Based on the detailed data of loan enterprises in a commercial bank system as training data, combined with the idea and method of data mining, the improved decision tree classification technology is applied to predict the loan risk of enterprises in commercial banks. The results show that the improved method can not only guarantee the accuracy of prediction rating, but also make the decision tree more concise, and the results can be used for reference.
【学位授予单位】：南京理工大学
【学位级别】：硕士
【学位授予年份】：2013
【分类号】：TP311.13;F832.4

【参考文献】