Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Optimizing Feature Ordering in Radar Charts for Multi-Profile Comparison

Albert Dorador

TL;DR

The paper tackles the problem that radar charts lose interpretability when axis ordering induces abrupt changes in polygon shape. It introduces a lexicographic minimax permutation strategy that exhaustively searches feature orders to minimize the worst-case mean adjacent difference across multiple profiles, using the maximum jump as a secondary tie-breaker. This approach yields globally optimal, reproducible orderings for moderate numbers of features while preserving the traditional radar encoding, contrasting with geometry-altering methods like OrigamiPlot and learning-based approaches like VON. The method enhances perceptual balance in multi-profile comparisons and provides clear guidance on computational limits and potential scalable extensions for larger feature sets.

Abstract

Radar charts are widely used to visualize multivariate data and compare multiple profiles across features. However, the visual clarity of radar charts can be severely compromised when feature values alternate drastically in magnitude around the circle, causing areas to collapse, which misrepresents relative differences. In the present work we introduce a permutation optimization strategy that reorders features to minimize polygon ``spikiness'' across multiple profiles simultaneously. The method is combinatorial (exhaustive search) for moderate numbers of features and uses a lexicographic minimax criterion that first considers overall smoothness (mean jump) and then the largest single jump as a tie-breaker. This preserves more global information and produces visually balanced arrangements. We discuss complexity, practical bounds, and relations to existing approaches that either change the visualization (e.g., OrigamiPlot) or learn orderings (e.g., Versatile Ordering Network). An example with two profiles and $p=6$ features (before/after ordering) illustrates the qualitative improvement. Keywords: data visualization, radar charts, combinatorial optimization, minimax optimization, feature ordering

Optimizing Feature Ordering in Radar Charts for Multi-Profile Comparison

TL;DR

The paper tackles the problem that radar charts lose interpretability when axis ordering induces abrupt changes in polygon shape. It introduces a lexicographic minimax permutation strategy that exhaustively searches feature orders to minimize the worst-case mean adjacent difference across multiple profiles, using the maximum jump as a secondary tie-breaker. This approach yields globally optimal, reproducible orderings for moderate numbers of features while preserving the traditional radar encoding, contrasting with geometry-altering methods like OrigamiPlot and learning-based approaches like VON. The method enhances perceptual balance in multi-profile comparisons and provides clear guidance on computational limits and potential scalable extensions for larger feature sets.

Abstract

Radar charts are widely used to visualize multivariate data and compare multiple profiles across features. However, the visual clarity of radar charts can be severely compromised when feature values alternate drastically in magnitude around the circle, causing areas to collapse, which misrepresents relative differences. In the present work we introduce a permutation optimization strategy that reorders features to minimize polygon ``spikiness'' across multiple profiles simultaneously. The method is combinatorial (exhaustive search) for moderate numbers of features and uses a lexicographic minimax criterion that first considers overall smoothness (mean jump) and then the largest single jump as a tie-breaker. This preserves more global information and produces visually balanced arrangements. We discuss complexity, practical bounds, and relations to existing approaches that either change the visualization (e.g., OrigamiPlot) or learn orderings (e.g., Versatile Ordering Network). An example with two profiles and features (before/after ordering) illustrates the qualitative improvement. Keywords: data visualization, radar charts, combinatorial optimization, minimax optimization, feature ordering
Paper Structure (12 sections, 4 equations, 1 figure)

This paper contains 12 sections, 4 equations, 1 figure.

Table of Contents

  1. Introduction
  2. Related Work: How Our Approach Fits in the Landscape
  3. OrigamiPlot and area-invariant visual transforms
  4. Learning to order: Versatile Ordering Network (VON)
  5. Our contribution
  6. Problem Formulation
  7. Proposed Strategy
  8. The minimax framework and its application to radar charts
  9. Complexity and practical recommendations
  10. Implementation and Example
  11. Limitations and Future Work
  12. Conclusion

Figures (1)

  • Figure 1: Radar charts before (left) and after (right) reordering. The optimized ordering yields a smoother and more balanced shape.