Research On Warning Information System Of Client Churn Based On Data Mining

Economic globalization causes the competition among the businesses fiercer andfiercer, so the enterprises in the information era should make use of the knowledgeconnoted in quantities of data to grasp the opportunities so as to enhance the corecompetitiveness. Clients are the key to success for the enterprises especially thecommunication enterprises. Applying data mining technology to customer relationshipmanagement (CRM) can provides quantification basis for the enterprises’ managementand decisions which can help the enterprises make an effective use of the limitedresources and increase the benefits in return. The prediction and control of the clientchurn is a challenge faced with the enterprises. Mass and frequent client churn prolongthe period of the profit recovery for the enterprises, which in turn causes great loss tothe enterprises. Most of the present studies on control of the client churn at home andabroad mainly focused on the aspects, like providing personalized service, analyzingthe clients’ satisfaction degree and loyalty index. Yet to test the effectiveness of themethods is difficult and they cannot solve the problems fundamentally.The thesis, on the basis of the existing problems of client churns in a mobilecommunication company in city C, mainly studies the application of the data mining toclient churn warning. This thesis builds up a client churn warning model based on SPSSClementine12.0and gives new thoughts for client churn warning.The main achievements of this thesis:(1) On the basis of studying and analyzing the principles of the client churnwarning model, this thesis puts forward the design object of the client churn warninginformation system. And then the thesis designs the concept model, system structureand function module for the system. (2) From the point of the system realization, the thesis implements and evaluatesits core part-the client churn warning model. SPSS Clementine12.0are used to set upthe client churn warning models, including C5.0node model, C&R node model,Logistic regression node model and neural net node model. By comparing the fourmodels, the thesis provides a more effective client churn warning for the enterprises.(3) Using.NET to call SPSS Clementine12.0connector, the thesis realizes thedynamic interacting between the data and rule sets of decision tree in the client churnwarning information system.