InsigHTable: Insight-Driven Hierarchical Table Visualization With Reinforcement Learning.

Area of Science:

• Data Visualization
• Human-Computer Interaction
• Machine Learning

Background:

• Hierarchical tables present complex data, but visualization can increase cognitive load.
• Existing methods for creating hierarchical table visualizations are often tedious and inefficient.

Purpose of the Study:

• To develop an efficient and insight-driven system for constructing hierarchical table visualizations.
• To address the challenges of a vast design space and tedious construction processes.

Main Methods:

• Propose InsigHTable, a mixed-initiative system integrating data insights and hierarchical structure.
• Utilize a deep reinforcement learning framework with an auxiliary rewards mechanism for sequential decision-making.
• Define data insights considering the hierarchical structure within table headers.

Main Results:

• InsigHTable effectively facilitates the construction of hierarchical table visualizations.
• The deep reinforcement learning framework with auxiliary rewards proves effective in uncovering data insights.
• Case studies and experiments validate the system's usability and effectiveness.

Conclusions:

• InsigHTable enhances the efficiency of creating hierarchical table visualizations.
• The system empowers users to better understand complex data and uncover hidden insights.
• Mixed-initiative and deep reinforcement learning approaches are valuable for data visualization tool development.