Web Services - Java Data Mining: Strategy, Standard, and Practice - page 273

has only the name of the task, it is executed asynchronously. With a
named task object, the client can inquire about the status of the task
execution by task name. Typically, short running operations such
as single record apply are executed synchronously and long-running
operations such as model building and batch apply are executed
asynchronously.
In addition to the executeTask operation, there are two other task-
related operations: getExecutionStatus and terminateTask that can be
used to monitor and control the execution of tasks in a data mining
engine (DME). The getExecutionStatus operation retrieves the status
of tasks that are currently submitted for execution, executing or
already completed. Since mining tasks can be long running, the ter-
minateTask operation is used to terminate task execution in the DME.
Chapters 8 and 9 discussed various types of mining objects: phys-
ical data, logical data, build settings, task, model, apply settings, and
test metrics. Similar to the Java API, JDMWS provides operations to
manage these objects. Table 11-1 describes all the mining object man-
agement operations that are used to save, retrieve, remove, and
rename mining objects.
The JDM standard allows vendors to provide a subset of the full
standard capabilities, while allowing them to extend those capabili-
ties. In the JDM API, appropriate supports capability methods are
defined in Factory objects so that applications can discover DME
capabilities at runtime. In the JDMWS, the getCapabilities operation is
used to retrieve the list of capabilities supported by a DME as a sin-
gle parsable XML document, which the requester can process to find
the capabilities of the DME. This operation is useful to customize
requests based on DME capabilities. For example, if the requester
prefers using the support vector machine (SVM) algorithm for classi-
fication and also wants to be interoperable with DMEs that do not
support SVM, it can have an internal rule such as “if the DME
supports SVM then use it, otherwise use the default classification
algorithm.” The request can then be customized based on supported
capabilities retrieved from the getCapabilities operation.

11.3.2 JDMWS Use Case

Chapter 7 described business problems for ABCBank and explained
how data mining techniques can be used to solve those problems.
This section extends the discussion by mapping these solutions to
JDM Web services and SOA design principles.