FP-Growth dalam Analisis Rekomendasi Barang Terkait

Abstract

Based on the implementation and analysis of the FP-Growth algorithm on retail transaction data, it can be concluded that this method is highly effective and efficient in discovering frequent item combinations within large-scale datasets through the FP-Tree structure, which compresses transaction records and enables pattern mining without generating candidate itemsets as required by the Apriori algorithm. This study involved several stages, including data preparation, preprocessing, FP-Tree construction, frequent itemset extraction, association rule generation, and evaluation. The experimental results show that FP-Growth successfully identified meaningful purchasing patterns, where the main frequent itemsets achieved support values ranging from 5% to 18%, indicating a significant level of occurrence in the transaction data. Furthermore, the generated association rules demonstrated confidence values between 70% and 92%, reflecting a high level of reliability in predicting the presence of related items when a particular item is purchased, while lift values greater than 1 confirmed positive correlations among item combinations. These results indicate that the discovered rules are both relevant and reliable for recommendation purposes. Overall, the findings highlight that FP-Growth is a powerful approach for uncovering hidden consumer purchasing behavior and can effectively support strategic decision-making in retail businesses, including product placement optimization, inventory management, cross-selling and up-selling strategies, promotional bundling, and the development of automated recommendation systems. In addition, this research provides opportunities for future work, such as applying the approach to larger datasets, integrating FP-Growth with machine learning techniques for adaptive recommendations, and implementing the method in real-time retail environments to enhance sales performance and customer satisfaction.