Analisis Perbandingan K-Means++, Mini Batch K-Means, dan Fuzzy C-Means pada Segmentasi Pelanggan

Abstract

Customer segmentation is a crucial process for optimizing marketing strategies This study aims to implement and compare three clustering algorithms on customer transaction data using RFMT (Recency, Frequency, Monetary, and Tenure) features. The dataset, obtained from the UCI Machine Learning Repository, underwent several preprocessing stages, including data cleaning, feature extraction, outlier handling, and normalization. Optimal cluster numbers were determined using the elbow method and validated using silhouette score and davies-bouldin index. The results show that mini batch k-means outperforms the other algorithms with the highest silhouette score of 0.4011 and the lowest davies-bouldin index of 0.9521. K-means++ demonstrated better computation time but slightly lower clustering quality, while fuzzy c-means produced less distinct segmentation