The Application Of Data Mining In Telecom Customer Churn Analysis

In a highly competitive market, customer churn in telecom has become a increasingly serious phenomenon.The problem is pressing, and is great significant for the telecommunications operations and management. If we can predict the loss of customers, analyse the reason and take retention measures, companies will be able to effectively prevent customer churn phenomenon.This paper studies how to use data mining to retain telecommunications customers. The main contents include how to predict the losses, and how to take retention measures. First, analyse the characteristics of churn, and use a multi-classification dynamic integrated model. Second, use clustering algorithm to establish a customer segmentation model, to find the most notable features of the lost customers, and take retention measures based on these features. The main findings are as follows:(1) proposed a complete solution framework. Including the forecast stage, analyse reason stage, calculate retain value stage, implement stage, evaluate and record stage.And analyzes the problems that enterprises should focus on each stage.(2) establish a churn prediction model for telecom customers, and the experiments show that multi-classification dynamic integrated model is better than a single classifier and multi-classification static integrated model to predict the loss of customers; also show that sensitive learning algorithm based on the cost is better than sampling algorithm on the prediction. And the influence in customer networks has a great impact on the prediction. Finally, the experiment compares the performance of time with a single classifier and static integratied algorithm.(3) proposed to analysis the reason of loss from the perspective of price and services, and establish a segmentation model for the loss, the VIP customers are divided into six groups, non VIP customers are divided into nine, summarize the features of each group, and develop retention tactics based on these features.