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

Concise representations of frequent high utility occupancy itemsets (FHUOIs), including maximal (MFHUOIs), closed (CFHUOIs), and generator (GFHUOIs) FHUOIs, play a crucial role in utility-driven mining. They offer significant advantages over the full set of FHUOIs, such as reduced size, improved efficiency, lower storage costs, and easier analysis. Moreover, MFHUOIs enable the recovery of all FHUOIs, while CFHUOIs and GFHUOIs together facilitate the generation of non-redundant high utility occupancy rules and the efficient retrieval of all FHUOIs and their key information. Despite their importance, existing methods mine CFHUOIs and GFHUOIs separately and lack an approach for mining MFHUOIs. To address this gap, this paper proposes two novel algorithms, namely MaxFHUOI-Miner and CGFHUOI-Miner. MaxFHUOI-Miner efficiently extracts only MFHUOIs using the extended pruning strategies EPNonMaxBr and LPNonMaxBr, which eliminate non-maximal FHUOIs early. Meanwhile, CGFHUOI-Miner simultaneously mines CFHUOIs and GFHUOIs using the novel pruning techniques EPNonCGBr and LPNonCGBr to quickly discard non-closed and non-generator FHUOIs at three levels of the prefix tree without subset relation 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.