and output objects required to execute tasks in the DME. For
example, the BuildTask object is used to build a mining model and
takes data specifications, build settings, and the output model
name, among others, as arguments.
Figure 8-12 depicts the JDM task-related objects. For each opera-
tion, we define a specialized task object. TestTask is the common
superinterface for testing supervised model quality. ClassificationTest-
Task is used to test classification models and RegressionTestTask is
used to test regression models.
The ApplyTask object is used for the apply operation. The
DataSetApplyTask is used to perform apply on a dataset with many
cases, whereas the RecordApplyTask is used to perform the apply
operation for single case, enabling real-time scoring. From a design
perspective, JDM places apply functionality in a separate package to
enable conformance for DMEs as scoring engines.
The ComputeStatisticsTask object is used to compute the attribute
univariate statistics that are discussed in Section 3.1.2. This task takes
as input physical data and an optional logical data and produces per
attribute univariate statistics. The logical data attribute types can
influence the type of univariate statistics computed by the DME.
ExportTask and ImportTask objects are used to perform mining
object import to and export from a DME. These tasks are particu-
larly useful to export models and settings from a data mining solu-
tion development system and import them to a production
Task objects and their relationships.