Credit Card Customer Churn Prediction With Binary Logistic Regression

Abstract

The increasing prevalence of credit card usage in Indonesia has brought significant benefits to the national economy but also presents challenges for the banking industry, particularly regarding customer churn. This research aims to analyze the factors influencing credit card customer churn using the Binary Logistic Regression method. Utilizing a dataset of 5,000 entries from Kaggle, the research incorporates demographic and behavioral variables such as age, marital status, product ownership, inactivity periods, and transaction patterns. Data preprocessing steps included handling missing values, encoding categorical variables, and feature scaling. The model was trained with 80% of the data and tested on 20%, with variable selection based on p-values (< 0.05). The results indicate that eight key factors significantly affect churn likelihood, including number of dependents, marital status, number of products, inactive months, number of contacts, total transactions, transaction count, and the ratio of Q4 to Q1 transaction amounts. The model achieved an accuracy of 87.10%, demonstrating strong predictive performance, especially for non-churn cases. These findings suggest that classical statistical approaches remain effective for churn prediction when supported by comprehensive data processing and relevant variable selection. The study contributes valuable insights for financial institutions to develop targeted retention strategies and enhance customer loyalty.