Discovering high utility itemsets based on the artificial bee colony algorithm

Abstract

Mining high utility itemsets (HUI) is an interesting research problem in data mining. Recently, evolutionary computation has attracted researchers’ attention, and based on the genetic algorithm and particle swarm optimization, new algorithms for mining HUIs have been proposed. In this paper, we propose a new algorithm called HUI mining based on the artificial bee colony algorithm (HUIM-ABC). In HUIM-ABC, a bitmap is used to transform the original database that represents a nectar source and three types of bee. In addition to an efficient bitwise operation and direct utility computation, a bitmap can also be used for search space pruning. Furthermore, the size of discovered itemsets is used to generate new nectar sources, which has a higher chance of producing HUIs than generating new nectar sources at random. Extensive tests show that the proposed algorithm outperforms existing state-of-the-art algorithms.