Analyzing the Effectiveness of Apriori and ECLAT Algorithms in Frequent Itemset Mining

Abstract

Consumer transaction data recorded through the point of sale (POS) system contains purchase patterns that can be used to support marketing strategies. In the context of convenience stores that have high transaction volumes and large product variations, product association analysis is important to uncover customers' shopping habits. This study compares the effectiveness of a priori and ECLAT algorithms in conducting frequent itemset mining on transaction data of daily necessities. Both algorithms were evaluated based on the number of rules generated, support value, confidence, lift, and execution time efficiency. The dataset used is Groceries, which represents actual transactions in the retail environment. The results showed that although a priori and ECLAT produced 25 rules. ECLAT was superior in execution times four times faster than Apriori without compromising the quality of the rules. Most rules have a confidence between 30–50% and a lift above 1.5, signifying a meaningful association. This study concludes that ECLAT is more suitable for use in complex and dynamic minimarket transaction data scenarios, and is recommended as the basis for the development of product recommendation systems and association-based bundling strategies.