Efficient approach for incremental high utility pattern mining with indexed list structure

Abstract

Since traditional frequent pattern mining approaches assume that all the items in binary databases have the same importance regardless of their own features, they have difficulty in satisfying requirements of real world applications such as finding patterns with high profits. High utility pattern mining was proposed to deal with such an issue, and various relevant works have been researched. There have been demands for efficient solutions to find interesting knowledge from specific environments in which data accumulates continuously with the passage of time such as social network service, wireless network sensor data, etc. Although several algorithms have been devised to mine high utility patterns from incremental databases, they still have performance limitations in the process of generating a large number of candidate patterns and identifying actually useful results from the found candidates. In order to solve the problems, we propose a new algorithm for mining high utility patterns from incremental databases. The newly proposed data structures in a list form and mining techniques allow our approach to extract high utility patterns without generating any candidates. In addition, we suggest restructuring and pruning techniques that can process incremental data more efficiently. Experimental results on various real and synthetic datasets demonstrate that the proposed algorithm outperforms state-of-the-art methods in terms of runtime, memory, and scalability.