Algorithm Research Of Cluster Analysis In Analytical Customer Relation Management

With the rapid development of scientific technology, the network technology and database technology have been applied widely and the volume of data stored by enterprises increases dramatically. The enterprises have a great deal of customer data and information, but they can not extract the business information contained in the mass data. How to find the rule and model from the data to acquire business information and help enterprises make strategic decisions better has become a hot topic of present researches. Customer Relationship Management (CRM) is a business strategy of enterprises, and customer subdivision is the principal issue of CRM. The application of data mining techniques to the customer subdivision can provide more valuable information for enterprises to carry out customer analysis.K-means algorithm is one of the essential clustering algorithms. It is a kind of clustering algorithm based on partitioning method. The thesis is planned to improve the algorithm based on the research while on the application aspect, the thesis take the algorithm into the customer segmentation use. Customer segmentation is the essential element for the enterprise to take out CRM. The major research work as follows:The thesis introduces the conception, function and flow of clustering technologies, and analyzes the advantages and disadvantages of common clustering algorithms and deeply discusses the K-means algorithm.Firstly, the rules of measurement functions are introduced against the simplicity of measurement function in the K-means algorithm so as to lead the algorithm to choose the related measurement functions according to different data sets. Secondly, the profile coefficients and the maximum distance function are adopted to determine K value and the initial cluster center so as to overcome the problems that the prior K-means algorithm is sensitive to the initial cluster center and the K value cannot be determined, and obtain a stable clustering result. The feasibility and effectiveness of the algorithm are confirmed by the result of Simulation experiments.Finally, the conceptions related to customer subdivision and subdivision methods are introduced. Meanwhile, both the traditional K-means algorithm and the improved K-means algorithm are applied to the customer subdivision. The comparison between the two algorithms'results proves that the improved K-means algorithm can effectively enhance the clustering effect and is more feasible in the practical application.