Targeted Querying of Closed High-Utility Itemsets
Shan Huang, Wensheng Gan, Jinbao Miao · 2023
In the era of big data, targeted querying of interesting itemsets having concise expressions is promising for improving the efficiency and capability of data mining applications. As the full set of high-utility itemsets is no longer explored, mining becomes more efficient. Nonetheless, identifying whether the current itemset includes the target pattern and is closed remains a challenge for enhancing data analysis efficiency. At present, no single-phase algorithm reliably identifies the targeted closed high-utility itemsets. In this article, we propose an algorithm called TQCUI, Targeted Querying of Closed high-Utility Itemsets containing the target patterns in a transactional database. The algorithm employs a compact attribute-utility-list structure for maintaining the utility and attribute information of the itemsets. Additionally, TQCUI utilizes several efficient pruning strategies to filter out unpromising itemsets, substantially reducing the search space. Moreover, to quickly prune non-closed itemsets, TQCUI introduces forward-extension and backward-extension checking schemes for the closure checking of itemsets. Extensive experimentation on both real and synthetic datasets demonstrates the TQCUI algorithm has good performance in terms of runtime, memory consumption, and scalability.