UNVEILING CHURN PREDICTION AT BANK IVORY

Abstract

The banking industry faces significant challenges in tackling customer churn within its credit card services. Customer churn refers to the situation where customers discontinue using a bank's services and migrate to another financial institution. To proactively address this critical issue, the present research endeavors to predict customer attrition in credit card services. To achieve this goal, the study extensively employs the CRISP-DM framework and diligently compares the performance of two predictive models, namely Gradient Boosting and Random Forest. The research endeavors to identify potential churn customers by analyzing crucial variables, including customer age, marital status, gender, income category, credit limit, and total transactions. The preferred modeling approach, determined based on the lowest misclassification rate, serves as a vital component of the research's analytical process. Remarkably, the research findings unequivocally demonstrate the superior performance of the Gradient Boosting model, which attains a misclassification rate of 0.1118 in predicting customer attrition.