<![CDATA[The hyper-competitive credit card industry faces growing challenges from digital disruption and evolving consumer expectations, demanding a shift from reactive to proactive customer retention strategies. Traditional reactive approaches prove ineffective as customer decisions often reach irreversible stages before intervention. This study aims to develop and evaluate a comprehensive data-driven framework for predicting customer churn at PT XYZ, a leading Indonesian banking institution, and design a scalable system architecture with CRM integration and real-time analytics dashboard for operational deployment. Following the CRISP-DM framework, we comparatively evaluate Logistic Regression, Decision Tree, and Random Forest using a dataset of 11,314 customer records. Model performance evaluation encompasses multiple metrics including Accuracy, Precision, Recall, F1-Score, AUC. Random Forest algorithm demonstrated superior performance, achieving an AUC of 0.98 and accuracy of 97 percent. Feature importance analysis revealed customer transaction inactivity and credit utilization patterns as the most critical churn predictors, with transaction count contributing 41.59% importance score. The research successfully establishes a robust foundation for data-driven customer retention strategies, providing PT XYZ with a comprehensive blueprint for institutionalizing proactive retention strategies that can minimize revenue losses and secure competitive advantages in an increasingly dynamic market environment.]]>
Copyrights © 2026
