Clutter Reduction Technique in Parallel Coordinate


#1

Guideline: Clutter Reduction in Parallel Coordinates
Visualization Using Axes Re-Ordering Based On
Minimal Edge Crossing

Source: http://ijcsit.com/docs/Volume%207/vol7issue5/ijcsit20160705033.pdf

Question:
Parallel coordinate plot is one of the most popular multidimensional data visualization techniques that suffers from disordered or crowded collection of graphical entities, as a visual clutter problem. I am currently developing parallel coordinate in high dimensional dataset, whereas clutter encapsulates somehow important patterns. One of the reduction technique according to Hemant Makwana et al, reordering the axes in Parallel Coordinates based on minimal edge crossing criteria and it optimises clutters in parallel coordinates.
I am wondering how to further improve the visual effectiveness of parallel coordinate plot, besides reordering with the edge crossing criteria. Also,color might be encouraged to reduce clutter combine with that technique. Any suggestions about different guidelines and techniques or reference paper from existing research could be quite helpful.
Thank you in advance!


#2

The State of the Art of Parallel Coordinates (2013) describes several techniques that may that may help:

  • each data-point does not necessarily need to be represented using a separate line: aggregation can be achieved by using polygonal envelopes, density contours, or continuous color-map representations of density

  • filtering can be used to reduce the number of lines that are rendered

  • axis scaling (including flipping) can make patterns in the data clearer

  • choosing a good axis ordering and/or allowing the viewer to interactively adjust the ordering can help (as you noted in your question)

In some cases it may also be clearer to draw multi-class parallel coordinates plots by drawing each class on a separate set of axes, rather than superimposing the lines for all datapoints on the same axes and distinguishing the classes using colour.