Design And Implementation Of Credit Card Analysis Data Mart

Through rapid development for years, China's credit card industry begins to take shape, and in pursuit of the development quality and the enterprise performance. It urgently develops the data analysis techniques to improve the scientific decision-making capability and the intelligent and automatic operation level. However, compared with the developed countries' counterpart industry, which has widely and deeply applied data analysis to all aspects of business, there is a big gap, especially in the data infrastructure and data mining capability.In response to these issues, this thesis reconstructs the Credit Card Management Information System in a large-scaled national bank. Based on the theory of data warehouse and data mart, the IT structure and the business of the bank, it develops the credit card analysis data marts depend on the Enterprise Data Warehouse. There into, the laboratory environment data mart for the data mining and ad-hoc analysis, and product environment data mart for the Online Analytical Processing and flexible reports.It designs 3 data tiers from the source systems to the users: Foundation, Aggregation and Application. Foundation is applied in EDW to store detail and historical data. Aggregation is used in data marts to support data mining and the Application, and the Application is directly used in reports and OLAP. Each tier has different structures - Normalization, De-normalization and Star Schema. The data is summarized and aggregated gradually. Each tier has different subjects and entities to match the data requirement properly.It is the first time for the bank having a credit card data analysis platform supporting data mining. The data marts integrate the data extracted from 6 credit card source systems: Credit Card System, Application System, Urgency System, Customer Services System, Card Point System and Acquire Business System. It makes the comprehensive analysis possible. The 3-tiers new Architecture: Source systems - Data warehouse - Data mart, not only avoids the inconsistency between CCM1S and EDW caused by data redundancy, but also balances the flexibility of the application and the data centralization.These data marts are successfully running in the bank under the conditions of huge amounts of data (10TB), meeting the data needs from analytical model development, flexible reports to multi-dimensional analysis. It becomes a solid foundation of the bank's credit card business.