A root node is the topmost node in a decision tree and the initial decision point from which every branch originates. It contains the first feature condition used to divide observations into mutually exclusive paths. section 6.3.6
Identify a root node by its structural and interpretive role:
- It appears at the top of a plotted tree and has arrows pointing toward the first internal nodes. section 6.3.6
- It establishes the first if-else split; subsequent branches apply additional feature conditions until reaching a leaf prediction. section 6.3.6
- Its feature is selected by finding the decision or split that best separates the class outcomes. section 6.3.6
First split in fourth-down conversion prediction
A root condition such as TO_GO <= 3.5 sends attempts into two branches. The resulting groups contain different counts of successful and failed conversions, after which internal nodes refine the prediction using features such as PLAY_TYPE and SCORE_DIFFERENTIAL. section 6.3.6
