Design of the JDM API - Java Data Mining: Strategy, Standard, and Practice - page 174

Each algorithm settings object is used to define algorithm-specific
parameters. For example, a TreeSettings object is used to specify deci-
sion tree algorithm—specific settings such as maximum depth of the
tree and maximum number of splits.

Cost Matrix
Section 7.1.4 discussed the use of a cost matrix for classification,
where a cost matrix is used to represent the costs associated with
incorrect predictions from a classification model. Since the cost
matrix supplements the build settings of a classification model, JDM
considers this object as one of the classification-specific settings
objects.
JDM defines a cost matrix as a named mining object that can be
managed separately by users and easily updated for model rebuild-
ing. A cost matrix can be specified for model build, apply, or test
operations depending on the capabilities of the particular JDM
implementation.
Figure 8-6 depicts the cost matrix object and its related objects.
Note that CostMatrix inherits from the CategoryMatrix , which repre-
sents any matrix that uses target values (a.k.a. categories) as row and
column header values. The CategorySet object, which is referenced by
the CategoryMatrix , contains the category values that are used to rep-
resent the matrix row and column values.

javax::datamining::data
CategorySet

1

1

javax::datamining::data
CategoryMatrix

javax::datamining
MiningObject

javax::datamining::supervised::classification
CostMatrix

Figure 8-6 Cost matrix interfaces.