Mining Discriminative High Utility Patterns

Abstract

Recently, many approaches for high utility pattern mining (HUPM) have been proposed, but most of them aim at mining high-utility patterns (HUPs) instead of frequent ones. The major drawback is that any combination of a low-utility item with a very high utility pattern is regarded as a HUP, even if this combination is infrequent and contains items that rarely co-occur. Thus, the HUIPM algorithm was proposed to derive high utility interesting patterns (HUIPs) with strong frequency affinity. However, it recursively constructs a series of conditional trees to produce candidates, and then derive the HUIPs. It is time-consuming and may lead to a combinatorial explosion. In this paper, a Fast algorithm for mining Discriminative High Utility Patterns with strong frequency affinity (FDHUP) is proposed by considering both the utility and frequency affinity constraints. Two compact structures named EI-table and FU-table, and two pruning strategies are designed to reduce the search space, and efficiently and effectively discover DHUPs. Experimental results show that the proposed FDHUP algorithm considerably outperforms the state-of-the-art HUIPM algorithm in all datasets.