The use of CRISP-DM methodology for data mining in banking

Peter Konda (2009) The use of CRISP-DM methodology for data mining in banking. EngD thesis.

PDF Download (2098Kb)

Abstract

Data mining has been recognized as an independent field of research for more than a decade. Introduced in 2000 CRISP-DM is considered the first formal methodology that fully covers the process of data mining. Large companies now seek to incorporate this technology into their existing systems. This thesis describes the uses of data mining in a bank. NLB, d. d. like most enterprizes in Slovenia established a data warehousing system. Using OLAP the employees can perform business analysis with ease, but may have problems finding complex patterns in the data. Therefore data mining represents a possible upgrade over existing systems. The first few chapters introduce data mining and its place in modern science. Since data mining deals with data I included a brief history of data storage development. The next chapters contain a full description of CRISP-DM methodology and techniques for solving common business problems. The research part covers the data mining process in practice. The objective is to calculate a propensity score for each customer. This was done iteratively using the SQL Server 2008 database platform with strong emphasis on data loading and analysis. I compared the accuracy of different classification models using graphic representation and cross-validation.

Actions (login required)