Optimizing Customer Purchase Insights: Apriori Algorithm for Effective Product Bundle Recommendations

Abstract

A retail store faces significant challenges in crafting effective sales strategies, particularly in designing promotional product bundles. To address this, the store leverages transaction data to analyze customer purchasing patterns, aiming to uncover products frequently bought together. This study employs data mining techniques, specifically the Apriori algorithm, to identify co-purchasing behaviors using 49,316 transaction records collected from January to June 2024. After thorough data cleaning and transformation, the Apriori algorithm identified 877 itemsets, spanning from frequent 1-itemsets to 4-itemsets. By setting a minimum support threshold of 0.003, the analysis narrowed down to 343 significant itemsets, including 325 frequent 1-itemsets and 18 frequent 2-itemsets, which served as the basis for generating association rules. Initially, 36 association rules were derived, highlighting various product relationships. To focus on impactful insights, the rules were filtered using a minimum confidence level of 0.5, yielding 3 highly relevant rules with lift ratios exceeding 1, indicating strong associations between antecedent and consequent products. These insights enable the store to design targeted promotional bundles, optimize product placement, and enhance overall sales performance. Additionally, this study demonstrates how data-driven strategies can provide a competitive edge by aligning with customer purchasing behaviors. To ensure continuous improvement, a Python-based system was developed, empowering the store to independently analyze transaction data and refine sales strategies in real time, adapting to evolving purchasing patterns as the dataset grows.