Database Reference
In-Depth Information
geneous actors with different roles, interests, levels of knowledge of the problem
domain and the territory.
Network and Graph Data:
Graphs appear as flexible and powerful math-
ematical tools to model real-life situations. They naturally map to transporta-
tion networks, electric power grids, and they are also used as artifacts to study
complex data such as observed interactions between people, or induced interac-
tions between various biological entities. Graphs are successful at turning seman-
tic proximity into topological connectivity, making it possible to address issues
based on algorithmics and combinatorial analysis.
Graphs appear as essential modeling and analytical objects, and as effective
visual analytics paradigms. Major research challenges are to produce scalable
analytical methods to identify key components both structurally and visually.
Efforts are needed to design process capable of dealing with large datasets while
producing readable and usable graphical representations, allowing proper user
interaction. Special efforts are required to deal with dynamically changing net-
works, in order to assess of structural changes at various scales.
4 The Visual Analytics Process
A number of systems for information visualization, as well as specific visual-
ization techniques, motivate their design choice from Shneiderman's celebrated
mantra “Overview first, Filter and zoom, Details on demand”. As is, the mantra
clearly emphasizes the role of visualization in the knowledge discovery process.
Recently, Keim adjusted the mantra to bring its focus toward Visual Analytics:
“Analyze first, Show the Important, Zoom, filter and analyze further, Details
on demand”. In other words, this mantra is calling for astute combinations of
analytical approaches together with advanced visualization techniques.
The computation of any visual representation and/or geometrical embedding
of large and complex datasets requires some analysis to start with. Many scalable
graph drawing algorithms try to take advantage of any knowledge on topology
to optimize the drawing in terms of readability. Other approaches offer repre-
sentations composed of visual abstractions of clusters to improve readability.
The challenge then is to try to come up with a representation that is as faithful
as possible to avoid introducing uncertainty. We must not fall into the naıve
assumption that visualization can offer a virgin view on the data: any represen-
tation will inevitably favor an interpretation over all possible ones. The solution
offered by Visual Analytics is then to let the user enter into a loop where data
can be interactively manipulated to help gain insight both on the data and the
representation itself.
The sense-making loop structures the whole knowledge discovery process
supported through Visual Analytics. A generic scenario can be given following a
schema developed by van Wijk [42], which furthermore admits to be evaluated
and measured in terms of eciency or knowledge gained. A choice for an initial
representation and adequate interactions can be made after applying different
Search WWH ::
Custom Search