A Unified Comparison of User Modeling Techniques for Predicting Data Interaction and Detecting Exploration Bias

Area of Science:

• Human-Computer Interaction
• Data Visualization
• Artificial Intelligence

Background:

• User modeling algorithms aim to understand user behavior in visual analytics for better data exploration.
• Existing algorithms can detect exploration biases and predict user interactions.
• A lack of rigorous comparison hinders the effective application of these user modeling techniques.

Purpose of the Study:

• To rigorously evaluate and compare eight distinct user modeling algorithms.
• To provide guidance on selecting appropriate algorithms for specific visual analytics tasks.
• To identify open challenges and future research directions in user interaction analysis.

Main Methods:

• Compared eight user modeling algorithms using a diverse set of four user study datasets.
• Analyzed algorithm performance based on exploration bias detection, data interaction prediction, and algorithmic complexity.
• Ranked algorithms based on their effectiveness across various metrics.

Main Results:

• Identified significant performance differences among the evaluated user modeling algorithms.
• Provided a comparative analysis highlighting the strengths and weaknesses of each technique.
• Established a ranking of algorithms based on empirical evidence.

Conclusions:

• The study fills a critical gap in the visual analytics community by offering a comparative framework for user modeling algorithms.
• Findings offer practical guidance for researchers and practitioners in choosing and applying user modeling techniques.
• Highlighted areas for future research include refining existing algorithms and developing novel approaches for analyzing visualization provenance and user interactions.