基于Clementine数据挖掘的银行信用风险精准度量

发布时间：2018-11-20 21:02

【摘要】：信用风险是商业银行的主要风险之一，根据麦肯锡的研究表明信用风险占总体银行风险的六成以上，比市场风险和操作风险之和的三倍还多。作为影响金融行业尤其是银行业兴衰荣辱的关键因素，信用风险历来是国内外学者和政府监管部门热点研究问题。随着全球化金融危机的影响以及信用交易的扩大化，商业银行面对的信用风险呈现形式多样化、操作复杂化的趋势，从我国的实际情况来看，信用风险度量的准确性与灵敏性都是制约银行业风险管理的薄弱环节，如何界定信用风险并施以有效的风险管理措施是当下商业银行追求有效利润并保证长久有效持续经营的重要问题，也是保障投资者利益、社会稳定的民生问题。本文从商业银行的角度出发，以信用风险度量指标体系的建立为核心，提出基于数据挖掘技术的商业银行中企业客户，尤其是上市企业客户的信用风险精准度量模型，希望能够为商业银行信用风险的度量提供一定的技术参考和方法依据。首先，论述了商业银行公司客户信用风险度量的相关理论，从巴塞尔协议出发结合当前国情对商业银行公司客户信用风险重新界定；介绍了信用风险精准度度量的模型和方法，研究了数据挖掘技术在商业银行中风险管理中的应用，尤其是在信用风险度量的必要性与可行性。其次，从指标体系建设和模型构建两方面详细介绍了商业银行信用风险精准度量的建模过程，系统分析了其影响因素包括财务因素、行业属性及宏观经济环境，得出了7个方面共40个指标构建的精准度量体系，并建立了以反向逐步选择法筛选变量、缩减指标数据的因子分析为基础的逻辑回归模型。最后，使用Clementine数据挖掘工具，根据选取的120个样本进行实证研究，利用其财务报表、行业属性及宏观经济数据，使用KMO检验、Bartlett球体检验的因子分析方法对指标进行降维，按照“CRISP-DM”数据挖掘流程建立了逻辑回归（LOGISTIC）模型。经过随机样本检测，，模型的准确性和稳定性较好，度量结果较为理想。结果表明，数据挖掘技术在商业银行企业客户信用风险度量中具有较好的预测效果。
[Abstract]:Credit risk is one of the main risks of commercial banks. According to McKinsey & Company's research, credit risk accounts for more than 60% of the total bank risk, more than three times the sum of market risk and operational risk. As a key factor affecting the rise and fall of banking industry, credit risk has always been a hot research issue for domestic and foreign scholars and government regulators. With the impact of the global financial crisis and the expansion of credit transactions, commercial banks face the trend of diversified forms of credit risk and complicated operation. The accuracy and sensitivity of credit risk measurement are the weak links of banking risk management. How to define credit risk and apply effective risk management measures is an important issue for commercial banks to pursue effective profits and ensure long-term and effective continuous operation. It is also a livelihood issue to protect the interests of investors and social stability. From the point of view of commercial banks and taking the establishment of credit risk measurement index system as the core, this paper puts forward the accurate credit risk measurement model of commercial banks based on data mining technology, especially the customers of listed enterprises. We hope to provide some technical reference and method basis for commercial bank credit risk measurement. Firstly, the paper discusses the theory of customer credit risk measurement of commercial bank company, and redefines the customer credit risk of commercial bank company based on the Basel Accord and the current situation. This paper introduces the model and method of credit risk precision measurement, and studies the application of data mining technology in commercial bank risk management, especially the necessity and feasibility of credit risk measurement. Secondly, the modeling process of accurate measurement of credit risk of commercial banks is introduced in detail from two aspects of index system construction and model construction, and the influencing factors, including financial factors, industry attributes and macroeconomic environment, are systematically analyzed. An accurate measurement system of 40 indexes in 7 aspects is obtained, and a logical regression model based on factor analysis of variable selection and reduction index data is established. Finally, using Clementine data mining tools, according to the selected 120 samples for empirical research, using its financial statements, industry attributes and macroeconomic data, using KMO test, Bartlett sphere test factor analysis method to reduce the index dimension. According to the data mining flow of "CRISP-DM", the logical regression (LOGISTIC) model is established. After random sample detection, the model has good accuracy and stability, and the measurement result is ideal. The results show that the data mining technology has a good prediction effect in the measurement of customer credit risk in commercial banks.
【学位授予单位】：天津大学
【学位级别】：硕士
【学位授予年份】：2012
【分类号】：TP311.13;F832.4

【参考文献】