| Data warehouse technology has been developed for many years and widely used in various industries,especially in commercial banks.Several large data warehouse software vendors in the international market,such as DB2,Oracle,Sql Server,etc,all provide a powerful data warehouse data technology products,so that the distributed business data sets can be stored in a unified storage with reliable hardware and software support.By using a data warehouse,commercial banks can greatly improve the efficiency of data usage.It lays a solid foundation for data analysis and data mining of various business systems within the enterprise.This paper is based on the data warehouse system project of a commercial bank data center in China.Firstly,the domestic and foreign development of data warehouse in recent years is introduced.Secondly,relevant theoretical concepts of data warehouse and data mining are introduced,which included data warehouse technology,the difference between data warehouse and database,data scheduling ETL(Extract-Transform-Load)system.By combining the characteristics of each business system of commercial bank,this paper analyzes the system requirements of commercial bank data warehouse projects,including data theme analysis,data storage planning analysis.The data warehouse architecture design is based on IBM BDW(Banking Data Warehouse)methodology,using IBM’s data warehouse software and hardware products to achieve.Through the development and design of the data scheduling ETL system,the data is extracted from each business data source system to the data warehouse for centralized storage.It also analyzes the business data and provides data analysis and display interface.The paper also uses machine learning algorithms to predict whether borrowers in the credit business will default on repayments.The data set with 10,000 defaults and 10,000 non-defaults is used as the research object.Data preprocessing is applied to the loan customer’s data,and machine learning algorithms such as logistic regression,support vector machine,random forest,K nearest neighbor and multi-layer perceptron are used to construct the default prediction model to predict whether the user defaults.The experimental results show that the multi-layer perceptron model has the best prediction effect.This paper studies the design of data warehouses in commercial bank and data mining,it constructs a data warehouse model which can effectively solve the data integration storage,realizes the centralized and unified storage of commercial bank business data,and provides a unified analysis platform for commercial bank business data analysis.By improving business philosophy,optimizing organizational structure,and supporting scientific and taking efficient data warehouse technology as a support,commercial banks can get better development in big data era.Figure 34 Table 13 Reference 57... |
