Mosaic Selections: Managing and Optimizing User Selections for Scalable Data Visualization Systems

Area of Science:

• Computer Science
• Data Visualization
• Human-Computer Interaction

Background:

• Interactive visualizations struggle with real-time analysis of large datasets (millions+ records).
• User selections for filtering data can be complex and require low-latency updates.

Purpose of the Study:

• To introduce Mosaic Selections, a novel model for managing and optimizing user selections in interactive visualizations.
• To enable efficient, real-time interaction with large datasets.

Main Methods:

• Developed Mosaic Selections, a model integrating filter predicates into data queries for visualizations and input widgets.
• Implemented automatic optimizations, including pre-aggregating data, based on query and selection predicate analysis.
• Formalized the selection model and optimization techniques within the open-source Mosaic architecture.

Main Results:

• Achieved orders-of-magnitude latency improvements for selection-based optimizations compared to unoptimized queries and existing Vega optimizers.
• Demonstrated efficient handling of complex, multi-component user selections.
• Validated scalability to millions and billions of records.

Conclusions:

• Mosaic Selections provide a flexible and interoperable framework for data filtering across visualizations.
• The model's automatic optimizations significantly enhance the performance of interactive visualizations with large datasets.
• Enables real-time interaction and analysis for massive data scales.