A Data-Driven Approach for Mapping Multivariate Data to Color

Shenghui Cheng, Wei Xu, Wen Zhong, Klaus MuellerarXiv 2016 · arXiv

Abstract

Visualizing high-dimensional multivariate data using color is a significant challenge in data analysis. Traditional methods often struggle to scale beyond three variables or fail to preserve the underlying relationships between data points, leading to ambiguous or "muddy" visualizations. This paper presents a data-driven approach that maps high-dimensional data into a 2D color space by leveraging Generalized Barycentric Coordinates (GBC) and the HSL color wheel. By incorporating PCA-based warping techniques, the system enhances visual contrast and ensures that similar data points are represented by perceptually similar colors, allowing for intuitive exploration of complex datasets.

Introduction

Color is one of the most powerful tools in data visualization, but its application is typically restricted to low-dimensional scalar fields. When dealing with multivariate datasets—such as environmental monitoring of multiple heavy metal pollutants—standard color blending often fails to scale or loses the underlying relationships between variables. The authors address this by creating a system that encodes high-dimensional similarities into a 2D color space (like a color wheel) so that similar data points end up with similar colors, and the dominant variables in a sample are visually obvious.

Methodology

The proposed pipeline transforms complex multivariate data into a readable "pseudo-colored" map through four primary stages:

  1. Similarity Analysis & GBC Layout: The system first calculates similarities between data items and variables. It uses an improved Generalized Barycentric Coordinates (GBC) plot to arrange variables (attributes) around the periphery of a circle. Data points are then positioned inside this circle based on their relative relationship to those variables.
  2. HSL Color Mapping: The circular GBC layout is directly mapped onto the HSL (Hue, Saturation, Lightness) color space. By fixing the Lightness (typically at L=0.65), the GBC plot aligns with the Hue-Saturation slice of the color wheel.
  3. Feature Extraction & Warping: Data points often cluster in small areas of the color space, leading to low contrast. The authors use Principal Component Analysis (PCA) to identify the distribution and offer three ways to "warp" the color space: Similarity Preservation, Contrast Enhancement (warping the data to fill the HSL space), and Comparison Compression.
  4. Continuous Map Generation: Finally, Adaptive Kernel Density Regression (AKDE) is used to turn discrete sensor points into a smooth, continuous spatial map.
Figure 2: Comparison of different color mapping modes. (a, e) Similarity Preservation, (b, h) Contrast Enhancement, (c, f) Compression, and (d, g) the original un-warped mapping. The warping techniques allow for better utilization of the color space to reveal hidden patterns.

Results and Interface

The approach was demonstrated using a dataset of 300 samples representing eight heavy metal pollutants (As, Cd, Cr, Cu, Hg, Ni, Pb, and Zn). The paper presents an interactive interface that allows users to explore this multivariate data through several integrated views.

Figure 1: The interactive visualization interface, featuring (a) parameter control, (b) the main pseudo-coloring plot, (c) the color legend, and (d) individual attribute scatter views.
  • Pseudo-Coloring Plot: A geo-map showing the combined multivariate data.
  • Color Legend Panel: A combined GBC/color map showing exactly where a specific color sits in relation to the eight pollutants.
  • Scatter View: Individual heat maps for each attribute, allowing users to verify the combined color against single-variable distributions.

Qualitative results demonstrate that the Contrast Enhancement scheme reveals fine-grained details in pollutant clusters that are otherwise invisible in standard, un-warped color mappings.

Discussion and Limitations

While the HSL-based GBC mapping effectively handles up to eight variables, scalability remains a concern as the circle periphery may become too crowded for human eyes to distinguish between adjacent variable colors. Additionally, the current implementation relies on a fixed Lightness value. Future work aims to transition to the HCL (Hue, Chroma, Luminance) space for improved perceptual uniformity and to better integrate lightness as a dynamic data-driven dimension.

Conclusion

This data-driven framework provides a robust method for multivariate color mapping. By treating the color wheel as a geometric coordinate system and applying PCA-driven warping, the system transforms complex chemical distributions into clear, interpretable visual patterns that preserve data similarity.

Summary updated

Abstract

A wide variety of color schemes have been devised for mapping scalar data to color. Some use the data value to index a color scale. Others assign colors to different, usually blended disjoint materials, to handle areas where materials overlap. A number of methods can map low-dimensional data to color, however, these methods do not scale to higher dimensional data. Likewise, schemes that take a more artistic approach through color mixing and the like also face limits when it comes to the number of variables they can encode. We address the challenge of mapping multivariate data to color and avoid these limitations at the same time. It is a data driven method, which first gauges the similarity of the attributes and then arranges them according to the periphery of a convex 2D color space, such as HSL. The color of a multivariate data sample is then obtained via generalized barycentric coordinate (GBC) interpolation.
BibTeX
@misc{cheng2016data,
  author = {Shenghui Cheng and Wei Xu and Wen Zhong and Klaus Mueller},
  title = {{A Data-Driven Approach for Mapping Multivariate Data to Color}},
  year = {2016},
  eprint = {1608.05772},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/1608.05772},
}
Figure