Mining of High Average-Utility Patterns with Item-Level Thresholds

Mining of High Average-Utility Patterns with Item-Level Thresholds

In this paper, we introduce a level-wise algorithm named High Average-Utility Itemset Mining with Multiple Minimum Average-Utility threshold (HAUIMMMAU), which relies on a novel transaction-maximum utility downward closure (TMUDC) property and a concept of least minimum average-utility (LMAU) to mine high average-utility itemsets (HAUIs). Two efficient strategies, named IEUCP and PBCS, are designed to further reduce the search space, and thus speed up the performance of HAUI mining. Several experiments carried out on both synthetic and real-life databases show that the proposed algorithm can efficiently discover the complete set of HAUIs while considering multiple minimum average-utility thresholds.