Abstract
Most traditional pattern mining is designed to process binary databases, so there is a limit to extracting meaningful information from real-world databases. To solve this problem, high utility pattern mining method for analyzing a non-binary database has been proposed and it is being actively studied recently. However, commonly encountered high utility pattern mining is not suitable for dynamic databases, which are subject to continuous changes, as they handle static databases. Transactions can be inserted, deleted, or modified from the database in a dynamic database environment. The pre-large method, which is one of the various techniques of processing a dynamic database, can efficiently operate by reducing the rescan of the original database using two thresholds. In this paper, we propose a method for mining high utility patterns applying a pre-large technique in an environment where transactions are continuously deleted. Also, we show on a basis of experimental results using real-world and synthetic datasets that the proposed algorithm exhibits better performance than the state-of-the-art methods.