The Applications Study Of Date Mining In Telecom Broadband Customers Churns Early-Warning

China's telecom industry is developing rapidly in recent years,but with the increasing of competitions,customers growth are becoming more and more slowly.And then,managers of telecom-company are troubled by the loss of customers.Customer is the source of profits,how to effectively manage customer churn has vital significance for the survival and development of company.The passing research only study customer chum from one-sided point of view,such as only from the point of customer dis-satisfaction or customer switching costs.An effective theory should be studied and used for customer churn forecasting as a support.In this article,the broadband customers of a telecom-company in he Pearl River Delta are studied,and Clementine data mining software used to establish a "input-transformationoutput" model,and C5.0,Logistics regression and neural network a(?)e used to training model respectively;and then the model are assessed and tested in practice;ultimately a suggestion for customer retention gotten.The main results of this study are as follows:First,the system theory is applied to the management of customer chum.According to this theory,the problem of customer chum is changed into non-steady-state of customer system,and that are an effective theoretical support for establishing an early-waming model.Second,an early warning model of customer chum based on system management theory are established,and it's an efficiently means for telecom operators to retain Customers.Third,the new regulation of customer-chum in the Pearl River Delta is funded.