Research Of Customer Churn Prediction In Telecom Industry Based On Data Mining

While computer and database are widely and commonly used in every aspect of the society, huge amounts of data has accumulated. Although the amount of the data is vast, the knowledge is poor. The importance of data mining which can change this situation has been recognized by data-intensive enterprise.In telecom industry a lot of adverse effects have been brought by the custom churn, and telecom operators need customer relationship management indeed. With the rapid development of warehouse and data mining, the author researched customer churn prediction in telecom industry based on data mining in the paper.The work belongs to the project which is named’Customer Churn Prediction System’of Q Mobile Communications Corporation. Data mining techniques is applied to predict the custom churn. In this paper the author designed and built integrated model using customer data as data set. SPSS is a kind of statistical analysis software and Clementine is a tool which is used for data mining. The system is built under the frame of business understanding, data understanding, data preparation, modeling, model evaluation and deployment.First, the theory of data mining and survival analysis was introduced, and classification algorithm and proportional hazards model was described in detail. Second, effective data cleaning, integration, conversion and exploration were made in process of modeling to improve the quality of the data. Also, the author balanced the unbalanced data sets and established the index system based on the experience in business and attribute reduction. Finally, the model which used a decision tree, neural network, logistic regression and Cox was built. And the author had an evaluation of the model. In order to maintain the customer the value of the customer was assessed. For different custom, different strategies were made. This can reduce the cost of retention and improve the rate of success.This paper combines the theory of data mining and the project. A model which can predict the churn custom is built. For theoretical research this paper can give a guide to build the classification model; for practice the result shows that the model can support decision and give corresponding solutions.