Customer Churn Prediction using Deep Learning

Abstract

Customer churn is a common challenge in the banking sector, severely affecting the profitability of financial institutions. Although various strategies have been implemented to address this issue, churn remains a persistent problem. To effectively mitigate this, the use of diverse predictive models is crucial. These models, built using machine learning and deep learning techniques, include methods such as classification, clustering, and hybrid approaches. The models like Artificial Neural Network, Convolutional Neural Network, Recurrent Neural Network, Deep Neural Network, Long Short-Term Memory are compared in this study. Across diverse models, DNN achieved the highest recall of 91%. Among the evaluation metrics, recall ensures to capture the most potential churners.