PERANCANGAN MODEL PREDIKSI CHURN NASABAH COMMERCIAL BANKING PADA BANK X DENGAN METODE DATA MINING

• Main Author: Martua Elkana, Audric
• Format: Final Project
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/84561
• Institution: Institut Teknologi Bandung

Product X is a digital banking service offered by Bank X that integrates various services such as Cash Management, Trade Finance, and Investment Services into a single platform designed for the convenience of business customers. Managed by the Transaction Banking Division, Product X had reached over 100,000 users by the fourth quarter of 2023. In a highly competitive market, with major competitors including large Indonesian banks, customer retention has become crucial due to the high costs associated with acquiring new customers compared to retaining existing ones. Bank X faces a significant challenge with customer churn, where customers close their accounts or allow them to become dormant. As of July 2023, the churn rate for Product X stood at 16%, which exceeds the banking industry's average of 10%. This highlights an urgent need for evaluating and enhancing customer retention strategies. A 5-Whys analysis identified the lack of a predictive churn model as the root cause, indicating the necessity for a data-driven approach to predict churn and develop preventive strategies. This study employs the CRISP-DM methodology to develop a churn prediction model by analyzing customer transaction data from January to July 2023. Four primary algorithms— Decision Tree, Random Forest, Adaptive Boosting, and XGBoost—were evaluated and refined through hyperparameter tuning. The XGBoost model demonstrated the best performance with an accuracy of 74.66%, precision of 81.51%, recall of 87.34%, and an F- score of 84.32%. Additionally, a prototype churn prediction system based on Streamlit was developed, enabling users to upload data, train the model, and predict churn. This research is expected to assist Bank X in enhancing customer retention and improving the efficiency of customer relationship management.