Beyond Frequency

Utility-oriented mining which integrates utility theory and data is a useful tool for understanding economic consumer behavior. Traditional algorithms high-utility patterns (HUPs) applies single/uniform minimum threshold (minutil) to obtain the set of HUPs, but in some real-life circumstances, specific products may bring lower utilities compared with others, their profit offer vital information. However, if minutil high, low are missed; low, number becomes unmanageable. In this paper, an efficient one-phase utility-oriented pattern algorithm, called HIMU, proposed HUPs varied item-specific utility. A novel tree structure multiple item set-enumeration (MIU-tree), global sorted conditional downward closure properties introduced HIMU. addition, we extended compact utility-list keep necessary information, thus HIMU model greatly reduces computational costs memory requirements. Moreover, two pruning strategies then enhance performance. We conducted extensive experiments several synthetic real-world datasets; results indicates that designed algorithm can address "rare problem" has better performance than state-of-the-art terms runtime, usage, scalability. Furthermore, enhanced outperform non-optimized approach.