An internal node is a nonterminal decision-tree node that tests a feature, applies a condition, and directs observations along subsequent branches. It lies between the root and leaf nodes, with both incoming and outgoing arrows. section 6.3.6 section 6.0
Interpret an internal node by following its decision rule:
- Evaluate the condition, commonly an if-else test on a feature value. section 6.3.6
- Follow the branch corresponding to the condition’s outcome. section 6.3.6
- Continue through later internal nodes until reaching a leaf node containing the final classification or regression value. section 6.3.6 section 6.3
Fourth-down conversion tree
A tree may first split on TO_GO <= 3.5, then use PLAY_TYPE or SCORE_DIFFERENTIAL at internal nodes to refine the prediction. For passing plays, a SCORE_DIFFERENTIAL threshold distinguishes successful from failed conversion predictions.
Loan approval
After the root tests whether age is 30 or fewer, an internal node tests income against $50,000. Its branches lead to approval or rejection leaves. section 6.3
