Predicting churn: how MultiLayer Perceptron method can help with customer retention in telecom industry

Abstract

Customer churn prediction has been used widely in various kind of domain especially subscription-basis industries. With the rapid growth of telecommunication industry over the last decade, this industry not only focuses on providing numerous products, but also satisfying the customers as it is one of the key solutions to remain competitive. This research proposed MultiLayer Perceptron Method for churn prediction. The evaluation is compared with three classifiers which includes are Support Vector Machine, Naïve Bayes and Decision Tree in term of several aspects. In preprocessing phase, we employed Principal Component Analysis and normalization to find the correlation among all the variables. For the postprocessing, InfoGainAttribute is used to identify the highest factor attribute that leads to customer retention. It is found that MultiLayer Perceptron outperforms other classifiers and international plan plays important role to retain customer from leaving organization.