Segmentasi Pelanggan dengan Model LRFM Menggunakan Particle Swarm Optimization dan Fuzzy C-Means

Abstract

Increasing business competition encourages companies to utilize e-commerce and better understand customer behavior through customer segmentation. This study combines the LRFM model with Fuzzy C-Means (FCM) and Particle Swarm Optimization (PSO), where PSO optimizes the initial cluster centers to improve FCM clustering performance. The research process includes LRFM calculation, data normalization, PSO parameter tuning, clustering, and evaluation using the Partition Coefficient (PC), Partition Entropy (PE), and Davies–Bouldin Index (DBI). The best PSO parameters were 50 particles, an inertia weight of 0.7, C₁ = 1.5, and C₂ = 1.0. The optimal result was obtained with two clusters, achieving PC = 0.845, PE = 0.267, and DBI = 0.613. The clustering identified lost customers with low activity and transaction value, recommended for a let-go strategy, and core customers with high activity and contribution, recommended for an enforced strategy.