The Effectiveness of Homogeneous Classifier Ensembles on Customer Churn Prediction in Banking, Insurance, and Telecommunication Sectors

Abstract

The prediction of customer churn is a big challenging problem for companies in different sectors such as banking, telecommunication, and insurance. It is a crucial estimation for many businesses since obtaining new customers frequently costs more than holding present ones. For this reason, analysts and researchers are focus on to investigate reasons behind of customer churn analyzing behaviors of them. In this paper, an ensemble-based framework is proposed to predict the customer churn in various sectors, namely banking, insurance, and telecommunication. To demonstrate the effectiveness of proposed ensemble framework, k-NN, logistic regression, naïve Bayes, support vector machine, decision tree, random forest, multilayer perceptron algorithms are employed. Moreover, the effects of the inclusion of feature extraction process are investigated. Experiment results indicate that that random forest algorithm is capable to predict churn customers with 89.93% of accuracy in banking, 95.90% of accuracy in telecommunication, and 77.53% of accuracy in insurance sectors when feature extraction procedure is carried out.