Vendor Implementations - Java Data Mining: Strategy, Standard, and Practice - page 409

includes online analytical processing (OLAP), and spatial and
statistics functions, also embedded in the database.
Oracle acquired Thinking Machines Corporation and its flagship
data mining tool, Darwin , in 1999, which began Oracle's investment
in data mining technology. Oracle Database 10g Release 2 (10.2)
includes all of the data mining functions supported under JDM 1.1,
and several of the JDM 1.1 algorithms.
This section explores some details of ODM features, the architecture
of the Oracle Java Data Mining (OJDM) implementation, JDM fea-
tures supported by Oracle, using the JDM application programming
interface (API) in the Oracle platform, Oracle-specific JDM extensions,
OJDM capabilities, and a brief discussion of ODM's SQL data mining
API, which is interoperable with OJDM-produced mining objects, and
the Oracle Data Miner graphical interface, which uses OJDM.

16.1.1 Oracle Position on JDM

Oracle is a strong supporter of open standards, in particular, the Java
standards developed under Java Community Process (JCP). The
JDM standard (JSR-73) was initiated and led by Oracle under the
JCP. ODM 10.2.0.1 supports JDM 1.0, while ODM 10.2.0.2 supports
JDM 1.1 [JDM11 2005]. Oracle continues its support and interest in
JDM as specification lead for JDM 2.0 (JDM 217).

16.1.2 Oracle JDM Implementation Architecture

OJDM is a thin-Java wrapper that conforms to the JDM standard and
enables Java applications to integrate in-database data mining features
of Oracle Database . OJDM uses the database as the DME and communi-
cates with the database using JDBC. OJDM supports the use of any type
of Oracle JDBC driver supplied with the database [ORAJDBC 2006].
Figure 16-1 shows the architecture of OJDM. Any type of Java applica-
tion that is compatible with J2SE 1.4.2 can use the OJDM API consisting
of the JDM standard API and Oracle-specific JDM API extensions. The
Oracle extensions are discussed further in Section 16.1.5.
OJDM maps the JDM named objects, such as mining model, task,
settings objects, and test metrics, to database objects as shown in
Table 16-1. The JDM mining model is mapped to the ODM mining
model database object. Mining tasks are mapped to database sched-
uler jobs that can be executed either synchronously or asynchro-
nously. Settings, cost matrix, and test metrics objects are mapped to