Analisis Perbandingan Metode K-Medoid dan Ward dalam Klasterisasi Nasabah Bank Churners

Abstract

This study compares the K-Medoid and Ward clustering methods in segmenting bank churner customers. The dataset used is the Bank Churners dataset, consisting of 5.000 observations and six variables related to credit card usage. The results show that the Ward method has a higher silhouette coefficient compared to K-Medoid, making it more effective in forming homogeneous clusters. Ward clustering produces four distinct customer segments, which can help banks develop targeted retention strategies. Based on the analysis, promotional strategies and personalized services are recommended to reduce customer churn according to each cluster’s characteristics.