The Churn Prediction Based On Customer Segmentation

With the economic globalization and the market internationalization, the competition between enterprises has intensified, and it leads to increasing loss of customers. Currently, the customer churn has become the main factors affecting business efficiency. Therefore, for many enterprises, how to maintain existing customers, identify potential loss of customers as early as possible and retain them in a timely manner, is the challenges that they currently faced and eager to solve urgently.The present paper analyzes the existing loss prediction model, and builds a loss rate prediction model that based on user segmentation. On the basis of IPTV data set, we put forward the multi-perspective and multi-dimensional (MPMD) user segmentation algorithm and the collaborative filtering (CF) prediction algorithm. The research mainly includes:(1) The paper improved the prediction model. In order to settle the inadequacy of the existing customer churn prediction model, we built a loss rate prediction model based on user segmentation, established its framework, and finished comparison experiment on IPTV data set.(2) Proposed the multi-perspective and multi-dimensional (MPMD) user segmentation algorithm. Based on features of IPTV data set, we analyzed the previous user segmentation method, and put forward MPMD user segmentation algorithm.(3) Proposed the collaborative filtering (CF) prediction algorithm. The existing lost rate prediction algorithm was only simply utilize user's property data or historical data, and did not consider calculating the lost rate by looking for the user's similar users. Therefore, we put forward the CF lost rate prediction algorithm:based on original lost rate predicting algorithm, looked for similar users, and then calculated a user's lost rate.Finally, through two groups of comparative experiments, it indicated the efficiency of model and algorithm that proposed in the paper.