Java Reference
In-Depth Information
To simplify API usage, OJDM infers logical data settings from
the data, and as such does not support the optional logical data
specification. Recall that logical data provides the renaming of the
physical attributes, attribute type specification, and attribute prop-
erties. In OJDM, attribute type is determined by the data type of the
table column: VARCHAR2 and CHAR data type columns are
inferred to be categorical attributes and NUMBER, FLOAT, and
INTEGER data type columns are inferred to be numerical
attributes. To change attribute type or rename attributes, a database
view definition can be used. For a more detailed list of supported
capabilities refer to Oracle Data Mining Application Developer's Guide
Oracle JDM Extensions
Oracle Data Mining has introduced several extensions to the JDM
standard API, as illustrated in Table 16-3. Package oracle.dmt.jdm is
the base package for the Oracle extensions to JDM. OJDM follows the
JDM standard framework for extensions. For example, OJDM intro-
duces the feature extraction mining function, where a feature represents
a combination of attributes that captures important characteristics of
the data. (See Chapter 18 for a brief description of feature extraction in
general, and Oracle Data Mining Concepts [ORADMCONCEPTS 2006]
for further details.)
Using MiningFunction.addExtension and MiningAlgorithm.addEx-
tension methods OJDM adds Oracle-specific mining functions and
algorithms, so that applications can view the Oracle-specific func-
tions from the standard interface. Feature extraction and other OJDM
extension classes/interfaces extend the relevant JDM base classes/
interfaces to be consistent with the standard. For more details, refer
to Oracle Data Mining Java API Reference [ORAJDMDOC 2006].
OJDM also provides interfaces for several of the data mining
transformations (e.g., binning, normalization, clipping, and text).
OJDM defines the API under the oracle.dmt.jdm.transform package.
OJDM also introduces new tasks to more fully automate the data
mining process. The high-level predict and explain tasks automate the
data mining process by hiding the complexities of attribute filtering,
data preparation, and the model build, test, and apply.
Search WWH ::

Custom Search