Application Of Data Mining In Analysis Of Customer Churn In Telecom Industry

As the open of telecom Industry, clients may have more choices to use different telecom products and services, and the competition between telecom corporations gets fiercer. Major telecom competitors have to face serous churn problems, while the telecom market is saturating. Various sales promotions are implemented to attract new customers in order to main market occupation. Investigation shows that the cost to attract a new customer is 5 to 10 times more expensive than to maintain a current customer, so a long-time loyal customer is more profitable than a new customer. In this case, how to maintain current customers and predict churn so as to adopt suitable marketing measures to prevent loss becomes a urgent problem for telecom operators.In this thesis, we apply data mining techniques to analyze churn problems in different views, utilizing history data from a telecom corporation. We use decision trees , neural network and logistic regression to predict churn. Then, cluster models are established to deal with different customer's groups in marketing, by using k-means algorithm in customers who have high value and high churn probability according to customer's behaviors.What we got through the research for the paper can be concluded in four points:1. Describe in details the process of establishing the churn models taking the data mining flow as the clue.2. Use three methods to select the candidate variable sets3. Discuss thoroughly the influence of result with different algorithms and model parameters in the process of establishing models.4. Give the detailed explanation and description to the results of clustering , had the very high practical value...