Leveraging Distance-Based Effectiveness Indicators for Enhanced Behavioral Pattern Discovery in Complex...

Area of Science:

• Cognitive Science
• Data Science
• Human-Computer Interaction

Background:

• Data-driven methods are crucial for process data analysis.
• Existing methods struggle with ambiguous actions in complex tasks.
• Interpreting behavioral patterns requires robust analytical tools.

Purpose of the Study:

• To introduce distance-based effectiveness indicators for analyzing actions in complex problem-solving.
• To enhance n-gram chi-square feature selection and K-medoids clustering with Dynamic Time Warping (DTW).
• To improve the interpretability and sensitivity of data-driven methods.

Main Methods:

• Utilized n-gram chi-square feature selection for action segment extraction.
• Employed K-medoids clustering with DTW distance for behavioral pattern identification.
• Introduced and applied state effectiveness (ds) and transition effectiveness (Δds→s') indicators.

Main Results:

• Effectiveness indicators significantly improved n-gram feature selection sensitivity.
• Clustering performance and interpretability were enhanced by effectiveness indicators.
• State effectiveness representations (ds→ds') demonstrated superior outcomes.

Conclusions:

• Effectiveness indicators offer a superior method for representing action sequences in complex problem-solving.
• Integrating effectiveness indicators into process analytics enhances the capture and explanation of behavioral patterns.
• This approach advances the analysis of collaborative problem-solving behaviors.