Segmentasi Pelanggan Berbasis Rfm dan K-Means pada Sistem Analitik Penjualan

Abstract

The growth of daily transaction data volumes in the retail sector often ends up as passive digital archives due to resource limitations in manual analysis. This condition makes it difficult for management to map consumer loyalty objectively, causing marketing strategies to miss the target. This study aims to design a web-based analytical information system using Python to automate customer segmentation. The approach used is the Knowledge Discovery in Database (KDD) cycle, integrating RFM (Recency, Frequency, Monetary) model feature extraction with the K-Means Clustering Unsupervised Machine Learning algorithm. From a total of 2,590 rows of raw transaction data, the system successfully pre-processed and aggregated them into 60 unique customer entities ready for clustering. Testing using the Elbow Method validated that the optimal number of partitions for this dataset is three clusters. Based on K-Means computation using Euclidean Distance, the system reached absolute convergence at the eighth iteration. This study successfully partitioned consumers into three segments: High Priority or Loyal Cluster, Medium Priority or Regular Cluster, and Low Priority or Passive Cluster. This analytical system is proven effective in visualizing segmentation autonomously, utilized as business intelligence to design targeted promotions and minimize customer churn risk. For future research, it is recommended to integrate directly into the point-of-sale system database in real-time and add customer demographic variables to make the analysis more comprehensive.

GS Cited Analysis