Performance Analysis of Association Rule Mining Algorithms: Evidence from the Retailing Industry

The Apriori and FP-growth algorithms have gained widespread popularity in various business applications. In the retailing industry they are widely used for market-basket data analysis and frequent pattern mining to gain valuable insights into customer purchasing behaviour. In this study, we conducted a comprehensive analysis of these two prominent association rules mining algorithms, utilizing six benchmark datasets from the UCI machine learning repository. Our investigation involved a thorough comparison of the execution time and the number of rules generated by both algorithms. Execution time is measured once by varying the support levels and next by varying the number of transactions and the support levels. Number of rules generated is estimated by varying the support levels of the rules. Through our rigorous experimentation, we derived insightful inferences that elucidated the utility of association rule mining in the retail industry. Moreover, we employed the Big-O method to compare the performance of the two algorithms and formulated a theorem that established FP-growth as Big-O of Apriori, substantiating the differences observed in their performance.