Subtree classification is a means to push classification from parent to child classes by defining reduced requirements for classification into the subtree.
Subtree classification uses the following additional values defined at the class level:
- A subtree threshold (60 percent by default)
- A subtree distance (10 percent by default.)
It is also possible to entirely exclude the parent class as valid classification result.
In the example below, without subtree classification, the parent class with 90 percent confidence would win. With subtree classification, classification would lead to the medium branch, and the child class with 60 percent confidence would win.
