Decision Tree Training Options

[General Options]   [Testing Mode]

The training mode is used to construct the decision tree. Records in the database are partitioned recursively into branches until nodes are formed that either (i) contain records that belong to the same class (or the majority in the same class) or (ii) too few records remain to perform any meaningful split.

Once the tree has been constructed, the model is stored in the project as an XML document. This document can be viewed after training or by editing the decision tree options and clicking View Tree. The model is saved to an external XML file when the project is saved (see decision tree node for details).

Min Objects In Node

The decision tree will stop splitting the number of cases if the number of objects in a node falls below this value. You can also use the spin control to set the minimum number of objects.

Class Dominance

The decision tree will stop splitting the data when the frequency of one of the classes in the node meets or exceeds this percentage. You can also use the spin control to set the class dominance value.

Max. Tree Depth

Select this option to stop building the tree after a certain level. For example, to avoid creating complex trees, leave this value at the default (10) in order to build the tree to a maximum depth of 10. Uncheck this option to build the tree to the greatest depth.

Visulization method

The method used to visualize the decision tree (standard, formatted XML or raw XML).