The Research Of Custom Churn Prediction Based On Data Mining Methods

In the earlier period of direct marketing, the income came mostly fromthe new customer. However, the market is more saturated today. This makes itmore difficult to obtain the newer and the cost is higher. So a lot of thecompanies cost more and more in reserving the customer. One of the ways toreserve the customer is taking some action before they get lost. And this is thevalue of the custom churn prediction.With the research object of data mining methods using in the customchurn prediction, the basis of the theories on CRM and DM, the paperanalyses the data of a telecommunication company using the tool of SAS. Thepaper makes a research of the whole proceeding, including the proceeding ofpreparing data, converting variables, building models and evaluating models.The paper builds one model for prediction and one for explanationaccording the task. The model for prediction provides a marketing list for theenterprise. The model for explanation helps the managers to understand thecustomer's behaviors. And both of the two models is readability in theirbuilding proceeding because the paper believes that the model is the assistantof the management.The paper tries to make a contribution in the actual use of data miningtechnology, and it also brings forward some new topics about the research,such as exploratory data visualization etc.