Cutting-Down the Complexity of Parallel Coordinates Plots in High-Dimensional Data

Adapted from: Johansson & Johansson. “Interactive dimensionality reduction through user-defined combinations of quality metrics.” Visualization and Computer Graphics, 2009.

Theoretical (Analytical):

Practical (Implementation):

Literature Work:

Overview and Problem Statement

Die usefulness and qualtiy of a parallel coordinates plot is highly influenced by the ordering of the dimensions. Putting correlated or similar dimensions next to each other, enables the user to identify interesting patterns, while noisy dimensions between patterns are counter-productive.

In datasets with a large number of dimensions, there is often not ‘one best ordering’ of the dimensions, but rather different useful orderings. Furthermore, some dimensions are not relevant for specific analysis tasks and make reading a visualization more complex.

The aim of this project is to optimize parallel coordinates plots by identifying relevant combinations of dimensions and searching for a useful orderings which highlight patterns for high-dimensional data.


  • Implement existing quality metrics for parallel coordinates plots in JavaScript.
  • Implement an algorithm which extracts subsets of dimensions for a specific task.
  • Optimize the ordering of these dimensions in a parallel coordinates plot.
  • Evaluation.


  • Good knowledge in information visualization
  • Good programming skills in Javascript


  • Scope: Bachelor/Master
  • 6 Month Project, 3 Month Thesis (Bachelor) / 6 Month Thesis (Master)
  • Start: immediately



  • State of the art of parallel coordinates [Heinrich and Weiskopf, 2013]
  • Pargnostics: Screen-space metrics for parallel coordinates [Dasgupta and Kosara, 2010]
  • Visualizing High-Dimensional Structures by Dimension Ordering and Filtering using Subspace Analysis [Ferdosi and Roerdink, 2011]