Bank Customer Churn Prediction with Machine Learning Methods

Abstract

This paper examines and analyses customer churn prediction in the banking sector using the data from ABC Bank. The analysis conducted will document the determinants of bank customer churn and provide insights to the most important factors which influence the customers decision to quit utilizing the services of a bank. The investigation is based on the results of two machine learning algorithms with k-fold-cross-validation and same boosting methods. The result of the analysis reveals that out of logistic regression and random forests algorithms, the random forest methods show a higher accuracy score which corresponds with the literature review studied. Furthermore, the statistic of the research indicates that customers age has the highest association with the likelihood of customer churning, while whether the customer has a credit card at the bank has the lowest interconnection. The results of this research may provide valid explanations to customer churn in the banking sector and bring further intuitions of the advantages which machine learning methods may provide to future financial analysis.