Scalability Analysis of Frequent Closed High Utility Itemset Mining on Multi-Year Retail Transaction Data

Frequent Closed High Utility Itemset Mining (FCHUIM) is a vital approach for discovering high-value patterns from transactional data. However, a major challenge arises as historical data volume grows substantially over time, particularly in dynamic retail domains. This study aims to analyze the scalability of the Closed-FHUIM algorithm with respect to increasing volumes of multi-year retail cooperative transaction data, spanning from one to five years. The evaluation focuses on four key performance metrics: execution time, memory usage, number of discovered patterns, and pattern growth rate. Experiments were conducted incrementally using annual transaction datasets. The results show that execution time grows exponentially with data volume, while the number of patterns increases significantly in the early years and plateaus in later periods. Memory usage exhibits fluctuating behavior influenced by transaction structures, and the pattern growth rate gradually declines as the data span widens. These findings suggest that although Closed-FHUIM is effective for high-utility pattern discovery, further optimization is required for deployment in large-scale and longitudinal retail scenarios.