Cyclic Polygon Plots
Maksim Schreck, Peter Albers, Filip Sadlo
TL;DR
The paper introduces cyclic polygon plots (CPP) to visualize $n$-dimensional data in 2D by decomposing into 2D subspaces and forming polygons that preserve dimension correspondence. It presents two cyclic-pair schemes, ab-bc and ab-cd, along with four placement strategies (intrinsic, geometric, angular, statistical) and benchmarks CPP against PCP and RC across multiple datasets, including a user study. Key contributions include the CPP formulation, intrinsic glyph placement, and a comparative evaluation showing CPP can outperform traditional methods in information density and readability, especially in higher dimensions, though challenges remain for value comparisons and identical vertices. The work demonstrates CPP's potential for scalable, quantitative multivariate analysis and offers practical guidance on scheme selection and placement, with log-scaling shown to enhance readability in dense data.
Abstract
In this paper, we introduce the cyclic polygon plot, a representation based on a novel projection concept for multi-dimensional values. Cyclic polygon plots combine the typically competing requirements of quantitativeness, image-space efficiency, and readability. Our approach is complemented with a placement strategy based on its intrinsic features, resulting in a dimensionality reduction strategy that is consistent with our overall concept. As a result, our approach combines advantages from dimensionality reduction techniques and quantitative plots, supporting a wide range of tasks in multi-dimensional data analysis. We examine and discuss the overall properties of our approach, and demonstrate its utility with a user study and selected examples.