Decision Tree Inference
The standard Decision Tree inference executes the Decision Tree (starting at the root node), executing the inference of the Objects on each Node it encounters and uses their Object value to determine which branch split to follow, until a Leaf Node is reached. the Decision Outcome of the reached leaf is the value assigned to the Decision Tree Object.
For Split node Branches, The engine executes the Object and uses it's value to match the condition in one of the branches to decide on which branch to follow:
- For single select List Objects: branches is based on Branch value membership
- For multi-select List Objects: branches is based on multi-select List Comparators and value sets
- For Numeric & Date Objects: branches is based on Numeric & Date Comparators and Thresholds (Fixed or Variable Thresholds)
- Otherwise Branch (optional): can be used as the last branch, if "none of the above" is true
- "Is Blank" Branch (optional): can be used as a branch that detects if the Object has no value (isEmpty=true)
If the node Leaf reached is set to Empty, then no Decision Outcome is assigned. In this case the isEmpty property and "Is Blank" status of the Decision Tree Object are set to true.
In the event a Sequence Node is encountered, inference will scan each of its branches in turn (top to bottom). The same above Decision Tree inference is then applied for each of the sequence node branches. a Sequence Node branch can be nested. A standard, single select decision tree will terminate when a non-Empty Leaf is encountered. a Multi-select Decision Tree continues inference until the a leaf in the final branch of the Sequence.
There are variations on the standard Tree inference, such as when Fuzzy Attributes are used in a Decision Tree.
