A Novel Algorithm for Mining Rare-Utility Itemsets in a Multi-Database Environment

Utility mining has recently been an emerging topic in the field of data mining. It finds out high-utility itemsets by considering both the important factors of profit and quantity. In some situations, rarely occurring items may co-occur in a relatively close relationship with specific high-utility items. These utility itemsets with rare items may provide useful information to decision makers as well. Most of the existing methods on utility mining were designed for a single database (centralized database) and not suitable for the environment with multiple data sources such as those in a chain-store enterprise. Moreover, the existing methods did not consider the existing periods and the existing branches of items. In this paper, we have thus proposed a new kind of patterns, named Rare Utility Itemsets, which consider not only individual profits and quantities but also common existing periods and branches of items in a multi-database environment. We have also proposed a new mining approach called TP-RUI-MD (Two-Phase Algorithm for Mining Rare Utility Itemsets in Multiple Databases) to efficiently discover rare utility itemsets. To our best knowledge, this is the first work on mining rare utility itemsets in a multi-database environment. At last, the proposed approach is shown to have good performance under a variety of system conditions through a series of experiments.