Efficient Mining of Concise Patterns for Fre quent High-Utility Occupancy Itemsets Using Extended Pruning Strategies

Tóm tắt

 Concise representations of frequent high utility occupancy itemsets (FHUOIs), including maximal FHUOIs, closed FHUOIs, and generators of FHUOIs, are essential in utility-driven pattern mining. These representations offer several advantages over the complete set of FHUOIs, such as reduced size, improved efficiency, lower storage costs, and eas ier analysis. Notably, maximal FHUOIs allow for the recovery of all FHUOIs, while closed FHUOIs and generators enable the generation of non-redundant high utility occupancy rules and the efficient reconstruction of all FHUOIs along with their key information. Despite their significance, existing methods either mine closed FHUOIs and generators separately or lack a solution for mining all maximal FHUOIs. To bridge this gap, this paper introduces two novel algorithms MaxFHUOI Miner and CGFHUOI-Miner. The former efficiently extracts only maximal FHUOIs using extended pruning strategies that eliminate non-maximal itemsets early, while the latter simul taneously mines closed FHUOIs and generators by employing innovative pruning techniques to eliminate non-closed and non-generator itemsets at three levels of the prefix tree without performing subset checks. Extensive experiments on real world and synthetic datasets demonstrate that the proposed algorithms outperform existing and baseline methods in both speed and memory efficiency, particularly for low minimum support and utility occupancy thresholds in dense databases.