Implementation of Apriori Algorithm to Analyze Sales Transaction Patterns in Official E-Commerce

Abstract

The growth of e-commerce in Indonesia has driven dynamic changes in consumer behavior, especially in the purchasing patterns of fashion products. However, most businesses have not optimally utilized transaction data to design targeted marketing strategies. One of the main challenges is the inability to systematically recognize customer purchase patterns from complex and large transaction data. This research aims to apply the Apriori algorithm, specifically the FP-Growth method, in identifying recurring purchase patterns based on product combinations that are often purchased together at the Qeela Official store, an e-commerce-based fashion business. The data used includes 20,000 transactions during the period January to April 2024, which were sampled into 10,000 transactions according to the RapidMiner system limitations. The research stages include data transformation to binary format, conversion of attributes to binominal, application of the FP-Growth algorithm, and formation of association rules using minimum support parameters of 0.001 and minimum confidence of 0.5. The results show the existence of strong association patterns, such as SHORT PARASUT → SHORT CARGO with a confidence of 82.1% and a lift of 3.197. The insights provide a strong basis for decision-making in product bundling strategies, cross-selling implementation, automated recommendations, and stock management. The data mining approach used is proven to be relevant and applicable to improve marketing effectiveness and operational efficiency in e-commerce businesses, especially in the highly competitive fashion industry.