Mining actionable combined high utility incremental and associated sequential patterns.

Area of Science:

• Data Mining
• Machine Learning
• Pattern Recognition

Background:

• High utility sequential pattern (HUSP) mining is crucial for actionable insights in various applications.
• Existing HUSP algorithms often generate redundant patterns by overlooking item/sequence relationships, limiting actionability.
• This redundancy compromises the performance of mined patterns in real-world scenarios.

Purpose of the Study:

• To introduce a novel method, Combined Utility-Association Sequential Pattern Mining (CUASPM), for HUSP mining.
• To effectively remove redundant patterns and extract high discriminative, strongly associated sequential pattern combinations with high utilities.
• To enhance the actionability and performance of mined patterns.

Main Methods:

• Incorporation of item/sequence relations into HUSP mining.
• Introduction of actionable combined mining concept.
• Development of a novel tree structure for discriminative HUSP selection.
• Implementation of global and local strategies for efficient HUSP mining with utility growth and high association.
• Introduction of two parameters to evaluate pattern interestingness for selecting actionable combined HUSPs (ACHUSPs).

Main Results:

• CUASPM effectively removes redundant sequential patterns.
• The method extracts high discriminative and strongly associated sequential pattern combinations with high utilities.
• Experimental results show CUASPM outperforms baseline methods in execution time and memory usage.
• CUASPM successfully mines high discriminative and strongly associated HUSPs.

Conclusions:

• CUASPM offers a significant improvement over existing HUSP mining algorithms.
• The proposed method enhances pattern actionability and discriminative power.
• CUASPM provides a more efficient and effective approach for discovering valuable sequential patterns in large-scale datasets.