<![CDATA[Retail transaction data are often used only for sales recording, leaving customer value and product affinity disconnected from managerial decisions. This study develops a web-based retail analytics framework that integrates RFM, K-Means, and global and segment-level FP-Growth to generate product-bundle, availability-check, and shelf-placement recommendations for SRC Inka Jaya. The analysis used 198,879 non-tobacco transaction lines representing 67,284 invoices, 2,565 products, and 1,257 identified customers from 1 January 2025 to 22 June 2026. RFM variables were transformed using , standardized, and evaluated for k=2–6. A four-cluster solution was selected as a trade-off between statistical quality and managerial usefulness, with a silhouette score of 0.371 and a Davies-Bouldin index of 0.934. Champion customers represented 28.48% of identified customers but contributed 54.35% of their monetary value. The primary FP-Growth scenario produced 963 rules, including 188 global and 775 segment-level rules, consolidated into 86 bundles, 46 availability-priority products, and 69 shelf-placement recommendations. Temporal validation achieved a 60.61% rule-match hit rate. Combining store-wide and segment-specific patterns produced more contextual recommendations and supported traceable delivery through an active-dataset dashboard.]]>
Copyrights © 2026
