CUSTOMER CHURN PREDICTION IN MOBILE BANKING APPLICATION BASED ON RANDOM FOREST

Abstract

The study was conducted to evaluate the performance of Random Forest in predicting customers’ tendency to abandon mobile banking apps and to explore the input features that contribute significantly to identifying customers’ intention to abandon. Qualitative, quantitative, and bibliometric methods were used in the study. The study used simulated data to train the Random Forest model with 5 input features, including (1) Login Frequency, (2) Balance Checks, (3) Transfer Transaction, (4) Online Savings, and (5) Bill Payments. Additionally, the study compared the performance of Random Forest with that of other supervised machine learning models, including Gradient Boosting, Logistic Regression, and SVM. The results show that Random Forest achieved the highest predictive performance, with up to 99.5% accuracy. Two characteristics were identified as strong indicators: “Login Frequency” and “Balance Checks.” From the research results, the application of supervised machine learning models in early identification of customers who are likely to leave the bank and perform periodic identification to ensure accuracy when customers change their consumer behavior, and business strategies need to prioritize focusing on customer groups that tend to leave, in addition, banks need to strengthen cooperation with businesses to create incentives to stimulate user demand.