Back to previous page
Conference Publication

Stable Search Radar - A flexible dimensionality reduction technique for mapping high-dimensional data to continuous 1d circular target spaces

Fraunhofer IGD, Germany
Hendrik Lücke-Tieke
Fraunhofer IGD, Germany
Tobias Mertz
MPE, Germany
Srividya Subramanian
Fraunhofer IGD, Germany
Thorsten May
Fraunhofer IGD, Germany
Jörn Kohlhammer
Jonas Stromberg
accepted in:
IEEE Visualization and Visual Analytics (VIS), 2023
Yet to be published
March 30, 2023
Read full article

Abstract—Radar visualizations lay out data in a circular pattern, allowing the center of the circle to act as an anchor. A possible usage scenario is decision-making on search results: The relevant results are close to the center while similar search results are located in proximity to each other at nearby angles [32]. When naively deriving the angle by reducing the high-dimensional input data to a 1d axis via dimensionality reduction, similar data items are mapped to similar angles. However, welding the axis’s upper and lower ends together to form a circle induces an error at that point. This bears the risk of users misinterpreting this boundary, hence only a part of the full 360 is usable for the data visualization. We prove the existence of this systematic mapping error around the welding point, show that circular dimensionality reduction techniques fail to create trustworthy circular projections directly, and propose the Cyclone mapping technique to overcome this. It is a data transformation approach that follows a two-step approach: First the high-dimensional data is projected to 2d via dimensionality reduction. Then, the data is mapped on a circle after placing a circle center in the 2d mapping. It therefore allows visualization designers to use the maximal screen space available for data ink while maintaining trustworthiness. In addition, we enable users to browse in an animated, visually stable fashion through the data landscape by interchanging the reference point located in the plot’s center. We evaluated the circle projection trustworthiness and the visual stability empirically. We also evaluated the usability, intelligibility and helpfulness of the Stable Search Radar in hallway testing, indicating that it enhances and supports users’ search behavior.

Download files:

Will be available soon to download

Get in touch

Thank you! Your submission has been received!
Oops! Something went wrong while submitting the form, try again please.