Machine Learning and Neural Network Models for Customer Churn Prediction in Banking and Telecom Sectors

Abstract

AbstractCustomer satisfaction is one of the most crucial elements which dictate the growth rate and success of companies. In the operational environment, customer churn is one of the most-ceiled challenges that make companies lose customers, and hence is of greater concern to the industries. Due to its straightforward impact on the revenues of the industries, they are trying to develop and adopt different ways to anticipate customer churn. Various factors that increase customer churning rates can hence be identified, and necessary actions can be taken to lessen it. This work aims to develop and analyze the churning rates in the banking department and the telecommunication company to foresee the customers who might churn. This work has used an open-source dataset of telecom and banking sectors’ customers and predicted churning rates using artificial neural networks (ANNs) and machine learning methodologies like decision tree, random forest, KNN, kernel SVM (K-SVM), naive Bayes, and logistic regression. The models are examined using different evaluation metrics, and the highest accuracy model is used for churn prediction. For the Bank dataset, random forest obtained the highest accuracy of 87.05%, and for the telecom dataset, artificial neural network obtained the highest accuracy of 81.93%. This work is based on the study of reasons for customer churning and methods to retain these customers. It will help in flourishing the level of performance and profit of the companies.KeywordsNeural networksCustomer churnChurn predictionMachine learningTelecomBank