Efficient algorithms for discovering high-utility patterns with strong frequency affinities

摘要

In recent years, high-utility pattern mining has been studied extensively. However, most of these studies have addressed mining high-utility patterns (HUPs) without consideration for their frequencies, leading to the mining of meaningless HUPs. One of the approaches to solving this problem is to use HUP mining with strong affinity frequencies. In this paper, we propose two algorithms to discover HUPs with strong affinity frequencies: DHUPMiner (Discriminative High-Utility pattern Miner) and its parallel version, DHUP-Miner*. Several novel pruning strategies are applied to reduce the search space for potential DHUPs. Experimental results show that the proposed algorithms are faster than the state-of-the-art algorithm (FDHUP) for both sparse and dense benchmark datasets. Moreover, the parallel algorithm (DHUP-Miner*) was found to handle large datasets well.