A Case Study for the Churn Prediction in Turksat Internet Service Subscription

Abstract

Churn prediction is a customer relationship process that predicts for customers who are at the brink of transferring all the business to competitor. It is predicted by modeling customer behaviors in order to extract patterns. An acquaintance of a customer is more costly than retainment of an existing customer. Churn predictions shed light on members about to leave the service and support promotion activities. These attempts are utilized to avoid subscription cancellation of existing customers. Nowadays, telecommunication companies take churn prediction very serious. They strive for monitoring customers in the business by using various applications in systematic approach. Our study is based on leading internet service providing company, Turksat Satellite Communications and Cable TV Operations Company's customer behavior analysis. It is the leading internet service provider of Turkey operating in telecommunications sector. We have created a two-phase solution utilizing data mining techniques. These are time series clustering and classification techniques.