Chui : Mining Closed High Utility Itemsets

Abstract

In association rule mining, a transaction is a set of items called itemset, where each item represents a product or a service that customer buy in one transaction. In an e-commerce application, an itemset represents a set of items that a customer bought in one transaction. Frequent Itemset Mining (FIM) is a very popular data mining approach which is essential to a wide range of applications. For a transactional database, FIM generates frequent itemsets i.e. groups of items (itemset) appearing frequently in transactions. However, one of the drawback of FIM is that it assumes that each item can appear only once in every transaction and that all items have the same importance (weight, unit profit or value). To address above mentioned issues, the High-Utility Itemset Mining (HUIM) has been defined. As opposed to FIM, HUI considers the case where items can appear any number of times in a transaction and where each item has a weight called utility (e.g. unit profit). Therefore, mining high utility itemset can be used to discover itemsets having a high-importance (e.g. high profit), that is called High-Utility Itemsets. An itemset is called high utility itemset (HUI) only if its utility is not less than a user-specified minimum utility threshold minutil. Discovering or generating high-utility itemsets in transactional databases is a popular data mining task. A limitation of traditional algorithms is that too many number of high-utility itemsets may be presented to the user out of which some are redundant. To provide a concise and lossless representation of results the support count measure can be considered, hence the concept of closed itemset mining can be used.