【作者】 李宇嘉；

【导师】 彭建刚；

【作者基本信息】 湖南大学 ， 金融学， 2005， 硕士

【摘要】 《巴塞尔新资本协议》内部评级法的实施对商业银行信用风险内部评级体系提出了最低要求，其目的是要求商业银行对信用风险进行有效细分和精确测算。这一目的的实现本质上是对银行内部信息系统的要求，即在信用风险管理工程化的趋势下，银行要凭借其内部信息系统强大的数据储存功能，通过建立评级模型和采用数据分析技术，实现客户信用评级的有效细分。这就为数据仓库(datawarehouse)的应用提供了契机。 数据仓库是支持管理决策过程的、面向主题的、集成的、随时间变化的、持久的数据集合。数据仓库所具有的海量数据储存功能、联机分析处理技术及数据挖掘(data mining)技术等这些强大的信息处理功能使数据仓库及其技术在信用风险内部评级方面的应用具有可行性。数据仓库及其技术在信用风险内部评级中应用的途径有很多：银行数据仓库的建立为内部评级提供了功能完善的基础数据平台；分类挖掘技术的应用可实现客户信用级别的有效细分；关联规则挖掘技术和时间序列分析在客户违约概率测算模型中得到应用。 银行怎样建立数据仓库?本文从内部评级的需求分析入手，先建立银行数据仓库的整体实施框架；然后从数据的处理、数据模型的建立、数据粒度的选择、数据集市等方面详细阐述；最后介绍了在建立数据仓库过程中银行应该注意的问题。客户信用评级是内部评级最重要的任务，在数据仓库建立的基础上，本文就神经网络分类挖掘技术在客户信用评级中的应用进行了举例分析。更多还原

【Abstract】 Implementing internal rating approach in The New Basel Capital Accord brings forward the least demands, the objective of these demands is commercial banks should subdivide efficiently and work out credit risk .Realization of this aim lean on information systems of commercial banks, that is based on the trend of engineeralization in management of credit risk, commercial banks must subdivide credit risk efficiently leaning on depositing information , setting up evaluation model, adopting data mining. It supplies chance for implementing data warehouse in credit risk internal rating.Data warehouse is data muster which sustains process of management and decision-making, faces theme, integrates all sorts of data, varies with time going and permanent for a long time. The use of data warehouse and technologies based on which in credit risk internal rating is feasible because data warehouse can deposit great capacity of data, and data warehouse technologies: On Line Analytical Process and data mining and so on take on powerful function in processing of information disposal. Data warehouse and technologies based on which can be used in credit risk internal rating from many routes: the setting up of data warehouse supplies basic data flat roof for internal rating; using data classification can realize the aim of subdividing credit risk efficiently; the mining of association rule and analysis of time-series and sequence can be used in models which are applied in working out probability of breach of faith.How do commercial banks set up data warehouse? This article starts with the analysis of the demands of internal rating, and founding total frame is the first step; Then article sets forth how to deal with data, set up data model, choose data granularity, found data mart; At last which problems we should pay attention to some problems related with the process of setting up data warehouse. Client credit rating is the most important mission. After setting up data warehouse of commercial banks, this article analyzes how neural network Model can be used in client credit rating by give an example.更多还原