Data Analytics: Exploiting the Data Warehouse - Data Warehouse Systems: Design and Implementation - page 347

Fig. 9.8 Excerpt of the decision tree for the Northwind case study
Given a decision tree model, a DMX content query over it can provide
statistics about the number of cases at each level of the tree or the rules
at each node, while a prediction query maps the model to new data in
order to produce recommendations and classifications. Metadata about the
model can also be retrieved. For example, over the model deployed (called
NWDW DecisionTree ), we can issue the following content query asking for the
characteristics of an interior node in the tree ( NODE TYPE =3):
SELECT NODE NAME AS [Node], NODE CAPTION AS [Caption],
NODE SUPPORT as [Support], [CHILDREN CARDINALITY] AS [Children]
FROM NWDW DecisionTree.CONTENT
WHERE NODE TYPE = 3
In this query, the attribute NODE SUPPORT contains the number of records
in the node, while the attribute CHILDREN CARDINALITY tells in how many
subgroups has the algorithm divided the node. This query results in the
following table. We can see that, for instance, the node labeled YearEstablished
> = 1975 and YearEstablished < 1990 contains 239 elements and has seven
children. On the other hand, the node Business Type = ' Delicatessen ' contains
43 elements and has two children.