Databases Reference
In-Depth Information
The calling communities are represented using a directed graph, in which the nodes represent the
subscribers whereas the links represent the calls. In particular, countries are mapped to unfilled cir-
cles, subscribers are represented by filled circles, the size and color of a filled circle reflect the total
number of calls made by the subscriber. The width and color of a link represent the overall level of
communication between the two ends. The tool also enables the user to drill-down on suspected
patterns. It should be pointed out that, in the real sense of the word, NicheWorks is not a data
mining system; it may be regarded as a visualization or exploratory tool. Therefore, the tool cannot
fully accommodate the entire mining process. Nonetheless, the tool is a classic example of the role
of visual data mining in visualizing raw data.
DBMiner
3
is an OLAP data mining system developed by the Data Mining Research Group
from the Intelligent Database Systems Research Laboratory at Simon Fraser University (
Han et al.
,
1996
). The system is owned by DBMiner Technology Inc
4
. It supports association rules, meta-
patterns, classification, and clustering. DBMiner provides a browser to visualize the OLAP process.
Association rules are visualized using bar charts and a three-dimensional ball graph view. With regard
to visualizing decision trees, the product offers a three-dimensional graph view and an optional grid
view. Clustering results are visualized using a two-dimensional graph view or a grid view. The user
interface is fairly simple and standard. However, it should be pointed out that users who are not
acquainted with data mining are likely to find the data mining environment somewhat intimidating.
It should be acknowledged that DBMiner does interface with MS-OLAP Server and also uses MS-
Excel as a browser for visualizing the OLAP process. Nonetheless, DBMiner provides no explicit
support for data/results export and import. Moreover, the effort does not report any evaluation on
the system.
KnowledgeSTUDIO
is a product by ANGOSS Software Corporation
5
. The product supports
decision trees, clustering, and neural networks. Decision trees can be constructed automatically
or interactively. The product relies heavily on standard business applications (e.g., MS Office) for
visualization functionalities. Due to its interactive and exploratory environment, data mining models
can be produced with relative ease. Consequently, KnowledgeSTUDIO has a short learning curve.
The product presently does not offer an explicit recourse for exporting data mining results (or for
exchanging data mining models in general). However, it is should observed that its support for
Predictive Model Markup Language (PMML) could facilitate export of results. There is no record
of any usability studies carried out on the product. The visual interface can be seen in Figure
3.21
.
Another effort relevant to the field of visual data mining is
VisMine
(
Hao et al.
,
1999a
,
b
) that
is based on the premise that new techniques for mining knowledge from large data warehouses often
exhibit the following problems: display problems (cluttered display and disjoint displays), limited
access, and lack of expandability. The main features of VisMine are: hiding non-primary structures
and relationships, unless the user focuses on them; supporting the simultaneous presentation of
multiple visual presentations ('slice and dice'); and providing architectural plug-in support to enable
