Churning of Bank Customers Using Supervised Learning

Abstract

In the current challenging era, there is prominent competition in bank industry. To improve quality and level of service, bank concentrates on customer retention as well as customer churning. This paper discusses the classification problem of banking industry. It focuses on the customers of a bank concerns towards churning, predicting the departing customers from potential customers. Machine learning is the cutting edge technology that is practical and handy to solve such problems. Using supervised machine learning, a proprietary algorithm (a typical machine learning model) is created to forecast and inform the bank about the customers who are at the highest risk in leaving the bank. A customer churn prediction can be used here as churn and nonchurn customers are to be defined. Using ML, gap is to be resolved between churn and nonchurn customers. Different accuracy levels are achieved by classifiers using different data sheets. A novel approach K-nearest neighbor algorithm (KNN) is presented in which dataset is suitably grouped into training and testing models depending on weighted scales along with XGBooster algorithm for high and improved accuracy.