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

Subspace analysis of high-dimensional data is extremely challenging due to the huge exploration space. We propose Subspace-Map, a novel approach with a map metaphor for interactive exploration of various subspaces. We utilize a subspace search algorithm to identify a moderate number of potentially valuable subspaces, each visualized as a city on the map. Similar cities are clustered into provinces and countries, highlighting common data and dimensional patterns that can guide users in constructing desired subspaces. With the map, users can grasp an overview of the exploration space and explore different subspaces via recommended tour routes in more detail. We demonstrate the effectiveness of Subspace-Map through cases with real-world data, experiments with user feedback, and a comparison with state-of-the-art subspace data visualizations.