ANALISIS PERBANDINGAN ALGORITMA K-MEANS DAN FUZZY C-MEANS DALAM SEGMENTASI POLA TRANSAKSI PELANGGAN E-COMMERCE BERDASARKAN MODEL RECENCY, FREQUENCY, AND MONETARY (RFM)

Abstract

The exponential growth of online transaction volumes across e-commerce retail platforms generates massive behavioral repositories, necessitating unsupervised machine learning to optimize customer relationship management. This study comparatively evaluates K-Means and Fuzzy C-Means (FCM) clustering algorithms in segmenting customer purchasing patterns governed by the Recency, Frequency, and Monetary (RFM) behavioural paradigm. A dataset comprising 4,372 customer transactions was preprocessed and standardized via Min-Max normalization. Optimal cluster cardinality was determined using the Elbow curve method and Silhouette Coefficient across k = 2 through 8. Experimental benchmarks indicate k = 4 as the most representative cluster partition. K-Means achieved an average Silhouette Score of 0.612 with a 0.084-second execution latency, outperforming Fuzzy C-Means which recorded a Silhouette Score of 0.587, 0.312-second runtime, and Partition Coefficient of 0.742. Consequently, K-Means delivers superior computational efficiency and well-defined cluster boundaries across Champions, Loyalists, At-Risk, and Churned segments to drive targeted e-commerce marketing interventions.