Telecom Churn Prediction Model Using Data Mining Techniques

Abstract

Recently several churn prediction models are being introduced that are able to predict possible churners to the service provider, in order to provide a retention solutions, i.e. offering them some compensations and special packages to keep them attached with their network. Predictive models can correctly identify possible churners. In this paper we are trying to implement a new prediction model that is uses data mining techniques on data obtained from a leading telecom company in Pakistan and then measure the generated benefits. The analysis shows a tremendous cost saving as compared to the expenses incurred in performing the churn management in a traditional way. We have employed the Decision Tree, Support Vector Machine and Neural Network classifiers on a six months subscriber’s data sample to predict the churners’ behavior.