Paper Info
Title
Understanding data dimensions by cluster visualization using edge bundling in parallel coordinates.
Abstract
This paper proposes an edge bundling approach to parallel coordinates to improve the visualization of data dimensions using cluster characteristics. The proposed edge bundling technique visually encodes the clusters information of each dimension, such as variance, means, and quartiles, into the curvature of lines. Each line is decomposed into multiple Bézier curves, in a way that instances belonging to the same cluster are bundled in a single curve, aiming to improve the user perception of the data. The hypothesis is that the proposed technique improves the interpretability of the data in each dimension and their relations between dimensions. Quantitative and qualitative tests were performed with participants to compare the proposed approach with the classical parallel coordinates and with another bundling technique. The results revealed that our approach outperformed in the majority of the tasks and confirmed the initial hypothesis, obtaining a low response time in average compared to other two methods, as well as having a positive aesthetic pleasing according to participants' opinion..
Year
DOI
Venue
2018
10.1145/3167132.3167203
SAC 2018: Symposium on Applied Computing Pau France April, 2018
Keywords
Field
DocType
Edge Bundling, Parallel Coordinates, Data Clustering
Cluster (physics),Interpretability,Curvature,Pattern recognition,Computer science,Visualization,Response time,Bézier curve,Parallel coordinates,Artificial intelligence,Cluster analysis
Conference
ISBN
Citations 
PageRank 
978-1-4503-5191-1
0
0.34
References 
Authors
13
5
Name
Order
Citations
PageRank
Rodrigo Santos Do Amor Divino Lima103.38
Carlos Gustavo Resque dos Santos226.46
Sandro De Paula Mendonça300.34
Jefferson Magalhaes de Morais412.72
Bianchi Serique Meiguins55628.03