Subspace-Map: Interactive Visual Analysis for Subspace Data with a Map Metaphor.

Area of Science:

• Data Science
• Information Visualization
• High-Dimensional Data Analysis

Background:

• Subspace analysis of high-dimensional data presents significant challenges due to vast exploration spaces.
• Existing visualization methods often struggle to provide an intuitive overview and facilitate targeted exploration.

Purpose of the Study:

• To introduce Subspace-Map, a novel interactive visualization approach for exploring subspaces in high-dimensional data.
• To enhance user understanding of data patterns and aid in the construction of meaningful subspaces.

Main Methods:

• Developed a map metaphor for visualizing subspaces, representing each as a 'city'.
• Employed a subspace search algorithm to identify key subspaces.
• Implemented clustering to group similar subspaces into 'provinces' and 'countries' for pattern highlighting.
• Designed interactive 'tour routes' for detailed subspace exploration.

Main Results:

• Subspace-Map effectively visualizes a moderate number of potentially valuable subspaces.
• The map metaphor and clustering highlight common data and dimensional patterns.
• Users can gain an overview of the exploration space and delve into specific subspaces.
• Demonstrated effectiveness through real-world data, user studies, and comparison with existing methods.

Conclusions:

• Subspace-Map provides an effective and intuitive approach for interactive subspace analysis of high-dimensional data.
• The visualization facilitates the discovery of data patterns and aids in guided subspace construction.
• Offers a significant improvement over state-of-the-art subspace data visualization techniques.