Implementation of Apriori and Fp-Growth Algorithms In Forming Association Patterns Based On Unwaha Cooperative Sales Transactions

Abstract

This study implements the Apriori and FP-Growth algorithms to identify association rules in sales transaction data from the UNWAHA Multi-Purpose Cooperative. Both algorithms successfully discovered product relationships, with similarities in rules for items like MAKARONI ASEP, KRUPUK PAK JONO, and KRUPUK 500. The FP-Growth algorithm, implemented using RapidMiner, outperformed Apriori in processing speed by 11 seconds and demonstrated higher accuracy in rule generation. Optimal results were achieved with minimum support and confidence values of 0.3 and 0.9 for Apriori (generating 5 rules), and 0.52 and 0.9 for FP-Growth (generating 6 rules). These settings balanced between generating too many rules, which could complicate interpretation, and too few, which might miss important patterns. Based on the analysis, strategic recommendations for the cooperative include implementing product bundling and discounts for frequently co-purchased items nearing expiration, optimizing product placement by grouping commonly associated items (e.g., MAKARONI ASEP, KRUPUK PAK JONO, KRUPUK 500, SOSIS SO NICE, and YUPI ALL VARIAN 5G) in easily accessible locations, and increasing stock for high-demand products like LE MINERALE. This research demonstrates the practical application of association rule mining in retail, offering data-driven insights to enhance sales strategies and inventory management for the UNWAHA Cooperative.