Design of the JDM API - Java Data Mining: Strategy, Standard, and Practice - page 170

object. Alternatively, during the modeling phase, a data miner may
have physical data that does not change until the solution is
deployed; however, the analyst changes the logical data to experi-
ment with various ways to interpret the logical characteristics of
attributes, in this case one physical dataset object with multiple logi-
cal data objects.
Figure 8-2 depicts the class diagram of the physical and logical
data-related objects and their relationships (objects with gray back-
grounds are enumerations). A PhysicalDataSet can have zero or
more physical attributes. This allows specifying only the dataset
URI without any physical attribute details and results in accepting
default characteristics for each physical attribute. Users can specify
physical attributes for a physical dataset to indicate per attribute
data type and role. In JDM the allowed data types and roles are rep-
resented as the enumerations AttributeDataType and Physical-
AttributeRole . The AttributeDataType has the enumerated values
double type, integer type, string type, and unknown type . The Physical-
AttributeRole defines the special roles some attributes may play in
the dataset. For example, the case id role is used to specify the
attribute that uniquely identifies each case in the dataset.
For the transactional data that was discussed in Section 7.4.3,
attributes and values are stored in multiple-records where the

javax::datamining
MiningObject

javax::datamining::data

AttributeType
categorical
notSpecified
numerical
ordinal

javax::datamining::data
PhysicalDataset

javax::datamining::data
LogicalData

javax::datamining::data

1

AttributeDataType
double Type
integerType
string Type
unknown Type

1 1

1

javax::datamining::data
DataPreparationStatus

1 n n 1

1

1

javax::datamining::data
PhysicalAttribute

javax::datamining::data
LogicalAttribute

prepared
unprepared

1

1

0..1

javax::datamining::data

javax::datamining::data
Attribute

PhysicalAttributeRole
attributeName
attributeValue
caseld
data
taxonomyparentld

1 javax::datamining::data

CategorySet

taxonomyChild

Figure 8-2 Physical and logical data objects and their relationship.