Authors:
Raphaël Tuor
1
;
Florian Evéquoz
2
and
Denis Lalanne
1
Affiliations:
1
University of Fribourg, Switzerland
;
2
University of Fribourg, University of Applied Sciences Western Switzerland and HES-SO Valais-Wallis, Switzerland
Keyword(s):
Visualization, Parallel Coordinates, Categorical Data, User Study.
Related
Ontology
Subjects/Areas/Topics:
Abstract Data Visualization
;
Computer Vision, Visualization and Computer Graphics
;
Databases and Visualization, Visual Data Mining
;
General Data Visualization
;
Scientific Visualization
;
Spatial Data Visualization
;
Visual Data Analysis and Knowledge Discovery
;
Visualization Algorithms and Technologies
Abstract:
Parallel Coordinates are a widely used visualization method for multivariate data analysis tasks. In this paper we discuss the techniques that aim to enhance the representation of categorical data in Parallel Coordinates. We propose Parallel Bubbles, a method that improves the graphical perception of categorical dimensions in Parallel Coordinates by adding a visual encoding of frequency. Our main contribution consists in a user study that compares the performance of three variants of Parallel Coordinates, with similarity and frequency tasks. We base our design choices on the literature review, and on the research guidelines provided by Johansson et al (2016). Parallel Bubbles are a good trade-off between Parallel Coordinates and Parallel Sets in terms of performance for both types of tasks. Adding a visual encoding of frequency leads to a significant difference in performance for a frequency-based task consisting in assessing the most represented category. This study is the first of
a series that will aim at testing the three visualization methods in tasks centered on the continuous axis, and where we assume that the performance of Parallel Sets will be worse.
(More)