Predicting Customer Churn in Telecommunication by Ensemble Learning

Abstract

AbstractRetaining the current customers to use the services is one of the most crucial tasks for a provider in order to increase profit. If a company has a good forecast about the customers who may change the service then the company can improve its services or propose a suitable service for customer. Hence, the company can successfully retain their current customers. Therefore, the customer churn prediction is very important for every provider. There has been a lot of work published recently for customer churn prediction. However, the prediction accuracy is still needed to improve. In this paper, we propose a method by combining multiple classifiers with weighted voting to form a robust model for predicting the customer churn. By conducting experiments on the telecommunication customer churn dataset, our proposed approach yields better results compared to the existing methods including Extreme Boosting, Random Forest, Support Vector Machine, Logistic Regression. KeywordsCustomer churn predictionsMachine learningEnsemble learningVoting