| With the fierce competition of the domestic telecommunication market,the pressure of operating the telecom companies is sharper and sharper.The telecom companies begin to realize the importance of the maintenance of the integration packages.This paper is based on the real need of the public maintenance in operating the telecom companies,aims to reduce the income loss.the market share cutting and other adverse effects caused by the churn customer.By using the data mining tools,this paper takes the study in the churn customer of the integration package of the telecom companies with the method combining theoretical analysis and empirical research。The main research contents are as follows:1,Requirement analysis:do in-depth understanding in the project and technical background,including the business definition of each component of the project,and the characteristics of the project and the churn customer.2,Overall item design:set up the experimental frame system about the churn customer project in integrated package,defines the project process,the frame of data mining,and the elements of the frame and relationship with the data flow between each other3,Data preprocessing:Based on the thorough understanding of the industry,effective operation in data clean,data selection,data integration and conversion,and data statute is undergoing,it contains:in the aspect of data clean,the key different data's clean work in daily management all the year round,the filling work of the null data of the ID card and not deleting the right data during the process of building the model to ensure that the data mining is in a right order;in the aspect of data conversion,complete the integration of the call accounts classified,derived for age,derived for the rate of the packages usage;in the aspect of data statute,complete the regulations of the abnormal attributes in call list and the packages usage;in the aspect of data integration,the final data extraction model is set by the advantage of the mining software which can be used as the relational view of data source,and the use of flexible deployment view.4,The mining model selected:various model of training and testing model are selected according to the project requirement and mining tools,and algorithm of input attributes and model parameters are selected by the use of SQL flexible Microsoft view.Estimating the result of the data mining,re-codified the input order in the worse one,the Naive Bayesian Model,and eventually forming a data mining model and thedata of production rules by improved much accuracy through the in-put cutting.5,Public maintenance deployment and execution:the reasons are summarized from the replies of the churn customer,and the customers who may or will loss are found by the model of data mining.The solutions how to maintain those customer with different three dimensional attributes are put forward to turn the crisis into opportunities and reduce the retention costs.Through the assessment of the implementation of the project,the data mining has reached the expected aim.In this paper,it combines the data mining theory with telecom enterprise operation management to establish a prediction model for churn customer of the broadband and Phone integration package and a process of public maintenance.It is showed that the data mining tools can guide and promote the public maintaining work effectively in the result of the project. |
