Diverse Interaction Recommendation for Public Users Exploring Multi-view Visualization using Deep Learning.

Area of Science:

• Human-Computer Interaction
• Data Visualization
• Artificial Intelligence

Background:

• Interactive visualization systems offer valuable insights but pose challenges for novice users in selecting appropriate interactions and data subsets.
• Decision-making for data exploration often requires professional expertise, creating a barrier for non-experts.

Purpose of the Study:

• To develop a method for providing diverse, insightful, and real-time interaction recommendations for novice users in multi-view visualization systems.
• To enhance user experience and data comprehension for individuals unfamiliar with complex data exploration.

Main Methods:

• Utilized a Long-Short Term Memory (LSTM) model to capture and encode user interactions and visual states into numerical vectors.
• Developed a recommendation system based on the encoded user behavior and system states.
• Evaluated the method using a visualization system focused on Chinese poets in a museum setting.

Main Results:

• The proposed method demonstrated workability in multi-view systems with diverse interaction types.
• User studies confirmed the model's effectiveness in guiding public users toward more insightful and varied interactive explorations.
• Participants using the system gained more accurate data insights compared to baseline exploration methods.

Conclusions:

• The LSTM-based recommendation system successfully addresses the challenge of guiding novice users in interactive visualization.
• The method enhances data exploration by providing tailored, real-time interaction suggestions, improving insight discovery for non-professionals.
• This approach has the potential to democratize access to complex data insights within interactive visualization environments.