with each function and algorithm typically provides one or more
supportsCapability methods to discover DME capabilities. For exam-
ple, to discover the capabilities of a decision tree implementation,
invoke the supportCapability(TreeCapability) method in the TreeSet-
tingsFactory object. Here, TreeCapability is the enumeration that lists
the decision tree capabilities supported by a DME.
Applications can also check the task capabilities supported by a
DME. Specifically, applications can discover whether a particular
task is supported for a specific function, algorithm, or for a function
and algorithm pair, for example whether the apply task is supported
for the k-means algorithm of the clustering function. JDM also
defines optional capabilities at each task level, for example, whether
the BuildTask supports specification of validation data that is used as
part of the model build operation.
The JDM standard defines Java interfaces to enable application
designers and developers to develop comprehensive data mining
solutions that are also portable across multiple DMEs. JDM objects
are primarily divided into data specification, settings, model, task
and test metrics. For each mining function and algorithm, we define
separate packages, which also contain related objects. Since vendors
will not typically implement all JDM features, conformance to the
standard is a la carte , allowing a vendor to choose among a subset of
features. JDM provides the ability to explore the capabilities of a
DME at runtime to implement portable code.
[Gamma+ 1994] Erich Gamma, Richard Helm, Ralph Johnson, John
Vlissides, Design Patterns: Elements of Reusable Object-Oriented Software ,
Reading, MA, Addison-Wesley Professional Computing Series, 1994.
[JDM11 2006] Java Data Mining 1.1 specification and API Java documen-
[JDMSpec] JDM 1.0 Standard Specification, http://www.jcp.org/
[Wikipedia-Confidence 2006] http://en.wikipedia.org/wiki/Confidence
[OMG_UML 2006] Unified Modeling Language, http://www.uml.org.