Performance Analysis of the Apriori Algorithm in Identifying Product Purchase Patterns in Transaction Data

Abstract

Republik Computer Slawi frequently faces inventory imbalances, characterized by stockouts for some products and overstocking for others. This study aims to analyze the performance of the Apriori algorithm in identifying product purchasing patterns to serve as a basis for data-driven decision-making. The analysis was conducted on 1,399 sales transaction records using RapidMiner, with a minimum support of 5% and a minimum confidence of 60%. The results demonstrate that the Apriori algorithm successfully generated frequent itemsets—up to 3-itemsets—and produced 10 association rules with confidence values ​​ranging from 0.855 to 0.981. The "Mouse" product exhibited the highest support value at 0.773, while the combination of "Mouse" and "Laptop Bag" showed the highest support among 2-itemsets at 0.662. Performance metrics yielded an RMSE of 0.060 and a Squared Error of 0.004, indicating a low level of prediction error for the model. Consequently, the Apriori algorithm proves effective in identifying purchasing patterns and can be utilized to support more optimal inventory management at Republik Computer Slawi.