Prediction of Customer Retention Rate Employing Machine Learning Techniques

Abstract

In the recent decade, telecom sector is subjected to major revolution in the terms of subscribers' count and the technological content. Owing to institutional and policy reforms that have liberalized the sector, there is a significant rise in competition among telecom companies. With companies experiencing an ever-increasing rate of customer churn rate, and as it is costlier to acquire a new customer than to retain an existing one, customer retention has become one of the prime focus of companies. Enterprises and managers are focusing on improving customer retention rate now more than ever. With the large amount of customer data at our disposal, machine learning algorithms are apt for the efficient analysis of such data. This study compares the accuracy of traditional data mining techniques viz Logistic Regression, Support Vector Machine (SVM), Decision Tree, XGBoost, Random Forest, Light Gradient Boosting, Gradient Descent Boosting and Cat Boost in predicting customer churn and then proposes an algorithm based on the above techniques that aims to find the major causes for customer churn from one company to another and the ways in which enterprises can improve customer retention. The reduced set of relevant features are selected by employing our proposed algorithm.