Investigation And Application Of Cluster Analysis In Customer Relationship Management

At era on" customer guide", Implementation of Customer Relationship Management management will set up a good last interdynamic relation with consumers. The purpose is " Sell products customer need really". However, different customers different demands. Only who explore customer's potential demand and offer different services and maintain the relation with the most valuable customer who creates profits and grows up . So, cluster and distinguish customers becomes primary problem to every manager .In the face of the information of more and more intricate customers, use the technology of Data Mining and Clustering will help enterprises to explore more information with commercial value which is favorable to marketing from the data piling up like a mountain .Utilize the analytical method of the cluster to find out relevant characteristic or mode from the data can extract its mode of consumers' behavior and distinguish the customers dynamiclly . And then offer individualized products. Which will obtain trust from customers. Finally making enterprises manage better.Because of insufficients in the traditional cluster method ,such as unusual number avulses of data, and kinds of level count difficult to define. The thesis has put forward a method of Restricted Minimum Variance Hierarchical Cluster. Its basic thoughts come from variance analysis . The variance is litter among similar cluster but heavy between different ones . Pretreatment the original data first , in order to weaken the impact on cluster course of unreasonable number value. Calculate variance between different samples or clusters. Amalgamate the small couple go on so .Calculate the test stone. Differentiate the number of cluster tentatively. Then choose principal components through scatter plot matrix . Determine to classify and count through synthesize comparative analysis. The ones that can directly perceived through the senses in the whole modeling course. The method can calculate many kinds of statistics technical indicator and look over the change of distance in same group or different groups. It can receive the result through the tree-likes figure and scatter plot.In order to verify models, the results is the same when input same variable with Quick Cluster . The characteristic is very obvious between different groups. It is different of purchase mode and consumption and behavior of different groups. It has proved that the original model is practical and has relatively strong convincingness.