Information Technology Reference
In-Depth Information
already been reused in systems built by others. The pluggability of the system
makes it easy to extend the functionality, to prototype new functionality, to
add more data sources to mash-up with previous data, etc. When no available
component does what the user wants, wrapping existing software is possible.
The biggest weakness of the system right now is that it is quite wasteful with
regards to memory. A lot of data is duplicated in different ways. Using XML for
the data also uses a lot of memory. We plan to extend the system to also allow
e.g comma separated lists of values instead of XML, though this will not help
with data being duplicated in both the system itself, in an adapter plugin, and
in the final visualization Webble, etc.
3 Other Similar Systems
Here we give a short overview of other systems for visualization and exploration
of data, and point out some differences between them and our system.
RapidMiner
[3] is an open-source prototyping system for knowledge discovery
and data mining. It is widely used and supports very many data mining and
machine learning algorithms. It is used to graphically set up work flows for data
mining. It supports multiple views of the same data and you can go back and
change some step of the work flow and get a different visualization results. Like in
our system, the underlying data format is hidden from the data mining operators
or visualization components, and changes made to a work flow are reflected in
all views of those data. The interaction when setting up work flows is a visual
process, but, there is no interaction with the visualized results.
Snap-Together Visualization
[4] (
Snap
) is very similar to our system. Different
visualization components are connected and e.g. selecting data in one compo-
nent is automatically reflected in connected components. As in our system the
interaction can be bidirectional so interactions with the second component are
reflected in the first component too. Components can be connected in different
ways, so selecting an item may select related items in one connected view, and
open details about the selected item in another view.
Snap
, like our system, has
a small interface that components need to follow and it is possible to wrap ex-
isting software with an interface wrapper (in for instance Visual Basic). Unlike
our system,
Snap
expects the data to come from a database (supporting ODBC)
and uses Microsoft's COM for component (process) communication.
The
VERD
system [5] uses IntelligentBox, another version of Meme Media,
for data visualization. It works with relational databases and has a wrapper
to treat Web resources as relational schema too. Using direct manipulation, it
allows interactive data exploration using various visualization methods. All vi-
sualization views set up are also treated as relations, making it possible to apply
the same operations to visualization views as to raw data.
VERD
requires the
IntelligentBox environment while our system runs in any Web browser, and in
VERD
the interaction only goes downstream in the visualization flow, i.e. inter-
action with one component does not affect components earlier in the flow.
