Exploratory Visual Sequence Mining Based on Pattern-Growth.

Area of Science:

• Data Mining
• Human-Computer Interaction
• Information Visualization

Background:

• Sequential pattern mining is computationally intensive and often yields uninteresting results.
• Existing methods struggle with large datasets and user-specific pattern relevance.
• A need exists for more focused and efficient pattern discovery approaches.

Purpose of the Study:

• To develop an interactive visual approach for sequential pattern mining.
• To enable users to guide pattern discovery and refine search spaces effectively.
• To create a "transparent box" model for understanding and controlling the mining process.

Main Methods:

• Integration of interactive visualization with pattern-growth algorithms.
• Introduction of local constraints to guide the mining process.
• Stepwise visualization of mined patterns for user feedback.

Main Results:

• Users can steer the mining algorithm towards relevant patterns.
• Local constraints significantly refine the search space without re-computation.
• Demonstrated effectiveness on web visit and activity sequence datasets.

Conclusions:

• Interactive visual sequence mining offers a more efficient and user-centric alternative.
• The proposed approach enhances the discovery of meaningful sequential patterns.
• This method improves the practical applicability of sequential pattern mining.