Towards efficient discovery of target High Utility Itemsets

Finding High Utility Itemsets (HUIs) in databases is crucial for identifying items that are of high importance (like profit) for decision-making. However, current High Utility ItemsetMining (HUIM) algorithms often ignore the interest or target of users in favor of effectively identifying categories of HUIs using various measures and constraints. As a result, these techniques usually return a set of HUIs that are either too large or small to employ in decision making. Nevertheless, it is apparent that users are often interested in a select group of HUIs which may be among the set of HUIs reported using existing techniques. While some recent and early works have offered methods for discovering user-targeted HUIs, these methods are neither memory- nor time-efficient as they depend on post-processing or pattern matching. Additionally, throughout the discovery process, these techniques are required to scan the database twice. To address these issues,this paper proposes an efficient Target High Utility Itemset Miner (TarHUIM). In contrast to current methods, TarHUIM uses the users’ target list and a single database scan to significantly reduce the search space and amount of time needed to find the user-targeted HUIs. Extensive experimental analysis show that TarHUIM is efficient and effective in discovering the set of targeted HUIs.