Design of the JDM API - Java Data Mining: Strategy, Standard, and Practice - page 179

javax::datamining::supervised
TestMetrics

javax::datamining::supervised::clasification
ClassificationTestMetrics

javax::datamining::supervised::regression
RegressionTestMetrics

1 1 1

0..1

javax::datamining::supervised::classification
ConfusionMatrix

0..1

javax::datamining::supervised::classification
Lift

0..1

javax::datamining::supervised::classification
ReceiverOperatingCharacterics

Figure 8-11 Test metrics objects.

Figure 8-11 depicts the class diagram of the test metrics object
relationships. There are two types of test metrics: one for classifica-
tion and one for regression. TestMetrics is the base interface. Classifi-
cationTestMetrics contains the subobjects ConfusionMatrix, Lift , and
ReceiverOperatingCharacterics . The RegressionTestMetrics object con-
tains the various error metrics mentioned in Section 7.2.5. JDM
defines test metrics as a mining object that can be saved with a name
by the DME. This allows applications to manage test metrics inde-
pendently from the models or data that produced them and to com-
pare test metrics across models and over time.

8.1.5 Tasks

JDM defines a Task object to support the JDM operations for model
build, test, apply, compute statistics, as well as import and export.
Tasks can be executed synchronously or asynchronously in the
DME. In general, data mining operations can be long running, for
example, when scoring millions of records or building models
using compute-intensive algorithms. Applications can execute such
long running operations asynchronously to avoid blocking the
application. The Task object encapsulates the specification of input