Java Reference
In-Depth Information
When a user community recognizes the need for a standard, it
will often look first to existing standards bodies before deciding to
form its own standards body. Often, it is more convenient to fit into
an existing framework with processes and recognition than to build
one from scratch. However, issues of control, effort, and timing can
lead a community away from existing standards bodies.
Some standards have strict guidelines concerning what can be
specified, how it must be specified, and how changes are allowed in
subsequent revisions. Others may have a rigorous process for sub-
mitting new material and revising existing material. This may
require travel to remote or exotic destinations for days or even weeks
as specifications are reviewed and edited. Time and resources are
required to provide material in the proper format, and to ensure
established procedures are followed.
In addition, the timing of when a standard becomes final can
affect the choice of a standards body. If the release cycle is several
years, a new user community may not be willing or able to wait for a
standard to be published, with the chance that it may not be
approved, requiring more editing and delay.
One of the challenges in having multiple standards in the same
domain is unifying terminology and semantics. A brief survey of
data mining literature will uncover numerous terms for the same
concepts, or slight variations on the same concept. For example, an
attribute in JDM can also be referred to as a predictor, variable, column,
feature, or field . In different contexts, each of these terms can also have
different semantics. For example, field could refer to a physical data-
base table column, or a logical input to model build or apply. Table
17-1 maps some data mining terminology for JDM, PMML, CWM,
and SQL/MM.
Some standards fit into an existing, highly evolved framework.
CWM is one such standard that has a fairly complex and comprehen-
sive data model. CWM/DM was required to fit within the overall
CWM framework.
Sometimes standards overlap in both concepts and terminology,
but have a different emphasis. For example, JDM focuses on execu-
tion of the data mining process including model build, test, and
apply, and mining object import and export. PMML, on the other
hand, focuses on the model representation and exchange using an
XML representation.
Search WWH ::

Custom Search