RASIPAM: Interactive Pattern Mining of Multivariate Event Sequences in Racket Sports

Area of Science:

• Sports Science
• Data Mining
• Human-Computer Interaction

Background:

• Racket sports analysis relies on tactical insights from competitor data.
• Data-driven pattern mining of event sequences is common but often yields tactics inconsistent with expert knowledge.
• This discrepancy can hinder expert understanding and application of data-derived tactics.

Purpose of the Study:

• To introduce RASIPAM, a novel interactive pattern mining system for racket sports.
• To enable domain experts to incorporate their knowledge into data mining for discovering meaningful tactics.
• To bridge the gap between data-driven insights and expert domain knowledge in sports tactics.

Main Methods:

• Development of RASIPAM (RAcket-Sports Interactive PAttern Mining system).
• Implementation of a constraint-based pattern mining algorithm responsive to expert input.
• Integration of a tailored visual interface for interactive comparison and adjustment of discovered tactics.
• Translation of expert suggestions in natural language into algorithm constraints.

Main Results:

• Quantitative experiments demonstrated the algorithm's support for real-time interaction.
• Case studies in tennis and badminton involving domain experts showcased the system's effectiveness and usefulness.
• The interactive workflow allows iterative refinement of tactics until expert satisfaction.

Conclusions:

• RASIPAM effectively integrates expert domain knowledge with data mining for racket sports tactical analysis.
• The system facilitates interactive discovery of meaningful and expert-aligned sports tactics.
• RASIPAM enhances the utility of data-driven methods by ensuring relevance to domain expertise.