Profit-Based HUI-Miner for Discovering Consumer Shopping Patterns in a Retail Store

Abstract

Retail transaction logs often reveal products that are frequently purchased together, but frequency alone does not show their economic contribution. This study applies a profit-oriented High Utility Itemset approach using HUI-Miner to identify valuable shopping patterns at YSL Grocery Store. The analysis followed CRISP-DM on 266,394 sales records from 2022. After attribute selection and missing-value removal, 251,225 records containing transaction, item, quantity, and price were processed. Item utility was represented by quantity multiplied by selling price, while the mining stage used a minimum utility of IDR 5,000,000, minimum support of 2%, and minimum confidence of 10%. Positive associations were retained when lift exceeded 1. The procedure produced 248 frequent itemsets, 77 confidence-qualified rules, 64 positive-lift rules, and 23 rules that satisfied all criteria. The strongest association linked two Sedaap instant-noodle variants with a lift of 3.26. The findings also show that support is not proportional to total utility: some cross-category combinations generated substantially greater utility despite lower occurrence. Therefore, retail decisions should combine utility, support, confidence, and lift when prioritizing shelf placement, bundles, promotions, and stock.