An Efficient Approach for Mining High Average-Utility Itemsets in Incremental Database
Ye‐In Chang, Chen Wu, Hsiang-En Kuo · 2025
Traditional high-utility itemset (HUI) mining methods tend to overestimate utility for long itemsets, leading to biased results. High average-utility itemset (HAUI) mining addresses this problem by normalizing utility with itemset length. However, uniform utility thresholds fail to account for varying item importance. Recently, HAUI mining with multiple minimum utility thresholds (MMU) has been used for flexible utility evaluation. While the generalized HAUIM (GHAUIM) algorithm performs well, it requires two database scans and is limited to static datasets. Therefore, we developed a novel tree-based method that scans the database only once to improve efficiency by reducing storage and eliminating costly join operations. Additionally, pruning strategies and incremental updates were introduced to enhance scalability. The developed method outperformed GHAIM in efficiency.