|
In China, since 2008 several telecom operators had been consolidated. As the loosening of the telecommunications industry control and new technologies applied, the wireless telecommunications market is saturated and fierce competition is increasing. Therefore, the customer churn prediction and management are even more important to telecom service providers. In such a saturated telecom market, they will want to meet the needs of customers and retain customers. Therefore, they hope that using technical method can predict the possible churners, then telecom service providers take a targeted strategy to these customers and attend to retain them. However, predicting the possible churners is difficult, and the dissertation use data mining technology to build a predictive model, find out the possible churners and provide personalized service.The main tasks of the dissertation are listed as below:1. Compare different data mining tools used currently and describe their advantages and disadvantages; introduce the current research of customer churn management .2. Establish relatively high prediction performance churn analysis model based on the decision tree and neural network. Give a framework of customer churn analysis system, and propose analytical procedure, mine raw data, then give the process of predictive model.3. Evaluate these models using empirical data. Explore data analysis, then extract the data, cluster customer, test models, verify its effectiveness and stability.Evaluation results show that using information of customers, contract, service status, call details and customer service related data, the model can effectively achieve accurate prediction data. As for the choice of different data mining techniques, we find that the decision tree or neural network applications can achieve good prediction accuracy.
|