Banks customer churn analysis and prediction using random forest

Abstract

Banks want to predict whether their customers will leave the banks in the future, which is called bank customer churn prediction because they want to survive in the market and maximize their profits. To accomplish that, previous studies tried many methods but none of them works very well. However, with the development of the internet, more and more data become available for banks to collect and use. To use these data efficiently, banks could use some machine learning algorithms to predict customer churn. Among many machine learning algorithms, random forest is one of the best and it was used in this work. The random forest model adopted 100 decision trees. It predicts the customer churn and computes the feature importance for all the variables included in the dataset. As a result, it could achieve an accuracy score of 0.875, which makes this model a promising method for predicting customer churn.