Strategi Retensi Nasabah Perbankan Lokal Berbasis Machine Learning: Analisis Perbandingan Algoritma Klasifikasi dan Teknik Resampling

Abstract

Customer retention has become an increasingly important strategic challenge for local banking institutions amid intensifying competition and the acceleration of digital transformation, making an understanding of customer loyalty patterns essential for designing effective and data-driven retention strategies. This study aims to analyze and compare the performance of machine learning algorithms in predicting customer loyalty in a local banking context, as well as to evaluate the impact of class imbalance handling techniques on model performance. Three classification algorithms—Decision Tree, Random Forest, and Logistic Regression—are employed in this study, with methodological stages including data preprocessing, model development, and performance evaluation. To address class imbalance in customer data, three approaches are applied, namely class weight adjustment, up sampling, and down sampling. Model performance is evaluated using the F1-Score and the Area Under the Receiver Operating Characteristic Curve (AUC-ROC). The results indicate that the Random Forest algorithm combined with the up sampling technique demonstrates the most consistent performance compared to the other algorithms tested, particularly in handling the minority class. The model achieves an F1-Score of 60% and an AUC-ROC value of 84%, indicating a good balance between precision and recall as well as adequate class discrimination capability. These findings suggest that ensemble-based machine learning models, supported by appropriate class imbalance handling techniques, can serve as effective decision-support tools for customer retention strategies in the context of local banking.