Database Reference
In-Depth Information
For instance, the
Relational
resource package, describing relational databases,
uses schemas (package) to structure the tables (classifiers). Every table (classifier),
in turn, contains a number of columns (features).
The diagram depicts that
Package, Schema
,and
RecordFile
are equal. In fact,
Schema
and
RecordFile
extend
Core'
s
Package.
The diagram also depicts that
Classifier, Table
,
RecordDef, Dimension,
and
ElementFile
are equal. Again,
Table
,
RecordDef, Dimension,
and
ElementFile
all extend
Core'
s
Classifier.
Finally, the
diagram depicts that
Feature, Column, Field, DimensionedObject,
and (XML)
Attribute
are equal. In fact,
Column, Field, DimensionedObject,
and (XML)
Attribute
all extend
Core'
s
Feature.
We have included this section in order to show how effective different applica-
tion types can be unified and modeled by elements of CWM. This approach is
widely used in XELOPES.
CWM and XELOPES
XELOPES is built on and compatible to the CWM standard 1.0. The
Data Mining
package of CWM is the central class extended by XELOPES. Moreover, meanwhile
XELOPES actively uses almost all CWM packages except for the
Warehouse
Management
layer.
In the next section, we focus on the functional description of XELOPES, and we
will systematically develop the XELOPES foundation. At this, we will briefly
explain which packages of CWM are used and which CWM classes are extended.
This is important because the structure of XELOPES is highly influenced by
the CWM.
We finally mention that the ordinary user of XELOPES does not need to know
much about CWM. In contrast, for users who extend XELOPES, it is helpful.
12.1.1.3 Business Intelligence Standards
XELOPES supports different standards from Business Intelligence. Beyond CWM,
these are JDM (Java Data Mining) [JDM] and JOLAP (Java OLAP) [JOLAP] and,
most importantly, PMML (Predictive Model Markup Language). PMML is a
standard for vendor-independent XML exchange of data mining models
[PMML]. PMML is supported by the core of XELOPES, and all of its models
can be exported into/imported from PMML. This applies not only to data mining
models but also to all agents, including that of recommendation engines. For this
purpose, the PMML standard was extended by prudsys for agents. A PMML file
contains all information required to apply an analysis model/agent like metadata of
the input, transformations, and the model itself. This makes this format very
compact and easy to use. We will not go into further detail here, since it is quite
technical. We just point out the PMML is used for serialization of all models and
agents of XELOPES.