Study On Cluster Analysis In Customer Relationship Management

The development of Information Technology brings the world' s economy into an increasing period that never appeared before.Customer Relationship Management, as the result of the development of information technology, has been the focus of enterprises' attention. It can provide customers' data and related data analysis, helping enterprises make full use of customer data to make business decision. It also provides scientific means and measure for enterprises to face competition. Lots of research and industry analysis show the fact that building and keeping customer relation is the only and most important base for competition advantage. And it' s also the direct result of the innovation of traditional business model, caused by new economy in our society.The core technology of customer relationship management is data mining. As a new technology to process business information, data mining can extract, transfer, analysis and modulate the mass data in the business database, to get the key data helpful to business decision and help enterprises to be intelligent in management.The tasks of data mining include association rules analysis, time series module, cluster analysis, classification and predication and so on. Cluster analysis, as a module and function of data mining, is the main content of this paper. This paper studied almost all the cluster algorithms to identify customer clusters. In our study, we value the algorithms' practical application instead of their complexity and perfection. In detail, this paper has done the following work:At first, we introduced the cluster theory in data mining and related statistical knowledge, and provided mathematical base for theintroduction of cluster algorithm.Second, on the base of lots of domestic and abroad reference, we did research on fuzzy c-means clustering algorithm, systematic clustering, subtractive clustering and so on. As for fuzzy c-means clustering algorithm, we introduced the concept of validity function to solve problems about partial optimization and how to decide the cluster number.Third, we borrowed the software of Matlab and SPSS and did experiment on a set of data. The experiment showed that the Matlab program was simple and the speed was quick so that it could be applied in large number of data. Our experiment got results of three clustering method and did comparison analysis.At last, the paper analyzed some problems related to cluster analysis, such as outlier, procession of odd data. They are in discussion now both on theory and practice. Besides, the paper analyzed some problems in the application of CRM system.Due to the condition and time limitation, our research is just the beginning of cluster analysis in CRM, which is to be developed in study and work later.