Interpretable Product Recommendation through Association Rule Mining: An Apriori-Based Analysis on Retail Transaction Data

Abstract

The rapid growth of e-commerce has generated vast amounts of transactional data, creating opportunities for data-driven decision-making in retail environments. This study presents an interpretable product recommendation approach based on association rule mining using the Apriori algorithm. Unlike complex black-box recommender models, the proposed method emphasizes transparency and explainability in identifying purchasing relationships. The Groceries dataset comprising 38,765 transactions was analyzed to discover frequent itemsets and generate actionable association rules. After applying minimum thresholds of 0.02 for support and 0.4 for confidence, a total of 67 frequent itemsets and 45 strong rules were obtained. The rule {whole milk, sausage, rolls/buns} → {yogurt} achieved the highest lift value of 1.66, revealing meaningful co-purchasing behavior. Visualization tools, including heatmaps and network graphs, were employed to illustrate rule strength and product interconnections, facilitating business interpretation. The findings demonstrate that interpretable rule-based recommendations can effectively support product bundling, cross-selling, and retail layout strategies. This study highlights the continuing relevance of Apriori in creating transparent, data-driven insights and proposes future integration with hybrid models to address personalization and scalability challenges in modern recommendation systems.