HUOPM: High-Utility Occupancy Pattern Mining.

Area of Science:

• Data Mining
• Database Systems
• Pattern Recognition

Background:

• Traditional pattern mining often relies solely on frequency, overlooking item importance and pattern representativeness.
• Frequent patterns can have low occupancy, limiting their applicability in real-world scenarios.
• Existing methods struggle to balance frequency, utility, and occupancy for comprehensive pattern discovery.

Purpose of the Study:

• To develop an efficient algorithm for mining high-quality patterns by incorporating utility and occupancy measures.
• To address the limitations of frequency-based pattern mining in transaction databases.
• To consider user preferences for frequency, utility, and occupancy in pattern extraction.

Main Methods:

• Proposes the High-Utility Occupancy Pattern Mining (HUOPM) algorithm.
• Introduces a novel frequency-utility tree and utility-occupancy list for efficient data handling.
• Employs pruning strategies based on global and partial downward closure properties to optimize search space.

Main Results:

• HUOPM efficiently discovers the complete set of high-quality patterns without candidate generation.
• Experimental results demonstrate the effectiveness and efficiency of HUOPM.
• The algorithm's derived patterns are intelligible, reasonable, and acceptable.

Conclusions:

• HUOPM offers a superior approach to pattern mining by integrating frequency, utility, and occupancy.
• The algorithm significantly outperforms state-of-the-art methods in terms of runtime and search space reduction.
• This work advances the field of pattern mining for practical, real-life applications.