《信用风险评估研究论文.doc》由会员分享，可在线阅读，更多相关《信用风险评估研究论文.doc（56页珍藏版）》请在金锄头文库上搜索。
1、基于贝叶斯方法的商业银行信用风险评估研究摘 要风险是亘古不变的研究课题,自从信用和金融诞生以来，风险就伴随其左右。每当人们自认为掌握了其中的道理的时候，它就以“经济危机”的形式给人类教训,2008年度的金融危机就是这种教训的典型代表。危机之后，各国政府便开始寻找其背后原因并实施应对措施。在2010年，G-20通过了新的巴塞尔协议即巴塞尔协议III，巴塞尔协议III它不仅加强了微观层面的监管，更加注重了宏观审慎的监管。巴塞尔协议III的产生反应了过去对金融风险监管的低效率事实，也说明了信用风险问题从来没有被彻彻底底的解决过。信用与违约这对孪生兄弟，在经历人类数百年的研究之后，依然充满了神秘，依然
2、有待深入研究。另外，我国“十二五”规划纲要中也明确强调，在新的五年里面我国要全面推动金融改革开放，构建多元、服务高效、监管审慎、风险可控的金融体系。商业银行在我国金融体系中占有举足轻重的地位，其信用风险状况关系到我国金融业整体的安全，研究商业银行信用风险具有重要现实意义。综上本文确定研究的对象：信用风险。伴随计算机技术的发展，贝叶斯统计也得到了快速发展，它逐渐成为统计学发展的前沿阵地,这是促使作者选择贝叶斯方法来研究信用风险问题的最初动力。另外，R、SAS、OpenBUGS、WinBUGS等软件的出现也使贝叶斯模型的实践成为可能,由此作者确定了完整的研究题目：基于贝叶斯方法的商业银行信用风险评
3、估研究。本文采用理论与实证相结合的方法，在搜集参考了大量资料的基础上，从商业银行交易对手（上市企业）的角度出发，研究不同贝叶斯信用风险度量模型在商业银行风险管理上的应用问题。研究过程中，我们选择商业银行的交易对手作为研究对象，一类是存在潜在风险的ST企业，另一类是相对安全的非ST企业；通过对企业的个财务指标作因子分析处理，最终选择能够反映指标大部分信息的6个主因子作为构建贝叶斯模型的影响变量；运用挑选出的敏感性主因子指标分别构建贝叶斯判别分析、贝叶斯Logistic回归模型、贝叶斯Poisson回归模型、贝叶斯二元分位数回归模型。实证结果表明，贝叶斯方法在信用风险度量上不仅具有传统风险度量模型
4、的准确性，还拥有更好的稳健性。关键词：信用风险；贝叶斯；判别分析；广义线性混合模型；二元分位数回归The research on credit risk assessment of commercial bank based on Bayesian methodsAbstractRisk was an everlasting topic and it was associated with credit. And whenever people thought that they knew its truth, it would punish humans in the form of eco
5、nomic crisis, and financial crisis in 2008 was a typical example. After the crisis, the governments around the world began to look for reasons and make some measures, and In 2010, the G-20 unanimously adopted the Basel III. The new basel III not only strengthened the micro-level supervision, but als
6、o paid more attention to macro-prudential supervision. And it also reacted that the supervision has low efficiency in the past, the credit risk issues had never been thoroughly solved. The twins of credit and risk were still full of mystery and they also needed to be researched in an experience of h
7、undreds of years. In addition, the outline of the 12th Five-Year Plan of China also explicitly stressed that we should promote the financial reforming and opening, and build a complete financial system which would have multivariable compositions, efficient service and prudential supervision. Commerc
8、ial bank occupied a pivotal position in Chinese financial system, and its risk would be related to the security of the whole financial industry, so it was of practical use to do research on the credit risk of commercial bank. Consequently we made the “credit risk” to be our research project. With th
9、e development of computer technology, the bayesian statistics also took a great step forward, and it had become the forefront of statistics. This phenomenon motivated me to study credit risk problem using bayesian statistics. And the emergence of statistical software such as R, SAS, OpenBUGS and Win
10、BUGS made the empirical analysis in this paper to be possible. Lastly the author made the decision that the research topic was “The research on credit risk assessment of commercial bank based on Bayesian methods”.This paper researched the commercial banks risk management problems by building differe
11、nt Bayesian credit risk measurement models, adopting the combination of theory and empirical methods and collecting a large amount of data and references, and the article perspective was on the basis of counterparties of commercial Banks (listed companies). On the research, we divided the counterpar
12、ties of commercial Banks into two categories, one had a potential risk marked as the STenterprises, the other one which was the relative safety flagged as non ST enterprises. With the help of factor analysis of seventeen financial indicators, we finally chose six sensitive main factors as the indepe
13、ndent variable in the bayesian models, and these factors could reflect most of the information of the indicators. Then the author built bayesian discriminant analysis, bayesian Logistic regression model, bayesian Poisson regression model and bayesian binary quantile regression model. The empirical r
14、esults showed that the bayesian method had not only the same accuracy as the traditional risk measurement model, but also has better robustness.Key words: credit risk; bayes; discriminant analysis; generalized linear mixed models; binary quantile regression目录论文总页数：55页第一章 绪论11.1课题背景11.2本课题研究的意义11.3研究
15、思路和研究框架21.4国内外研究现状41.4.1国外研究状况41.4.2国内研究状况51.4.3贝叶斯方法在信用风险中运用的研究概况6第二章 基于贝叶斯方法的信用风险评估理论基础82.1信用风险评估理论基础82.1.1信用风险定义及诱因82.1.2信用风险的特征92.1.3信用风险的度量92.2贝叶斯统计理论基础132.2.1条件概率132.2.2先验分布、似然函数与后验分布142.2.3贝叶斯定理152.2.4贝叶斯计算16第三章 基于贝叶斯方法的信用风险评估模型203.1贝叶斯判别信用评估模型203.2信用风险的贝叶斯广义线性混合模型213.2.1贝叶斯Logistic回归信用风险评估模型
16、213.2.2信用风险的贝叶斯广义线性混合模型223.3基于贝叶斯方法的分位数回归风险评估模型253.3.1分位数回归253.3.2二元分位数回归253.4模型优越性诊断26第四章 基于贝叶斯方法的信用风险评估的实证研究284.1样本数据的说明284.1.1数据来源284.1.2指标体系的建立284.1.3指标数据的处理304.1.4指标的简化314.2基于贝叶斯方法的信用风险评估的实证结果354.2.1贝叶斯判别实证结果354.2.2贝叶斯广义线性混合模型的信用风险度量结果364.2.3基于贝叶斯分位数回归模型的信用风险度量结果394.3小结40第五章 结论与展望425.1总结425.1不足与展望43参考文献45附录48附录A 数据处
