INFORMATION VISUALIZATION ON SPHERES - Information Visualisation: Techniques, Usability and Evaluation - page 36

2. Computational Performance: Force-directed layout algorithms
iterate until the level of energy remaining in the system converges
to a threshold. The sphere's topology may benefit force directed
convergence as items can “wrap-around” the sphere. The layout
algorithm was executed 20 times for each of the datasets in both
spherical layout and a flat layout, with the same settings for
convergence in each. The spherical layout achieved convergence
33% faster on average with the Twitter dataset, but there was no
significant performance improvement with the stock dataset (Fig.
2.13). This may indicate that a spherical layout may be able to
converge more quickly or to potentially result in more optimal
layouts. However, this is inconclusive as performance may be
susceptible to the nature of the data and the parameters of the force-
directed layout models.
Fig. 2.13. Performance comparison of flat and spherical force-directed graph
layouts with two different datasets.
User Feedback
SphereCorr was evaluated in informal interviews with ten expert traders
and professional portfolio managers from capital markets with varying
degrees of familiarity of visualization ranging from expert to neophyte. A
preamble discussing the tasks associated with correlations, representations
the users may have seen before, and the suggestion that securities could be
placed on a sphere immediately preceded a short demonstration and/or use
session.
All users were familiar with viewing correlations in a numerical
matrix. Only three users were familiar with either a coloured-matrix or a
flat projection. None had seen visualization of a graph on a sphere before.
Almost universally there was an immediate visceral response from
expert financial practitioners, sometimes expressed verbally, such as