Comparative Analysis of Association Rule Mining Algorithms: An Application to Grocery Store

Abstract

Market Basket Analysis (MBA) is a vital retail data mining technique that reveals patterns in customer transaction data, identifying products frequently purchased together, enabling retailers to optimize sales strategies, enhance customer experiences, and streamline inventory and marketing efforts. Using a dataset from Ilorin Shoprite, this study explores purchasing patterns and their applications in retail, marketing, and bioinformatics for informed decision-making. The study analyzes grocery transaction data to uncover significant product associations, aiding strategic decisions like catalog design and cross-selling, and compares three association rule mining algorithms Apriori, ECLAT, and FP-Growth evaluating their time efficiency, memory usage, and effectiveness across small and large datasets, with the goal of identifying the best algorithm for retail transaction data and providing actionable insights for businesses. Data preprocessing involved a structured grocery dataset, followed by applying Apriori, ECLAT, and FP-Growth to identify purchase associations, where analysis of 1000 unique transactions revealed key associations such as WHT BREAD SWEET and EVA WATER 750ML. FP-Growth outperformed others in speed for both small and large datasets, while Apriori and ECLAT excelled in generating frequent item sets, making FP-Growth recommended for large-scale MBA due to its efficiency. The identified associations offer practical insights for refining catalog design and cross-selling strategies, and retailers should adopt FP-Growth to optimize MBA performance, leveraging purchasing patterns to enhance marketing and customer experiences across industries like retail and bioinformatics.