Decision Tree
Last Updated: July 29, 2026 | By Mihail Sebastian | AI Dictionary
A supervised model that predicts by asking a sequence of yes/no questions about the data, with every step readable by a human: the most interpretable ML model.
What is a Decision Tree?
A decision tree is a supervised learning model that predicts an outcome by asking a sequence of questions about the data, one question per node, until it arrives at an answer at a leaf. It handles both classification (categories) and regression (numbers).
Its distinguishing property is interpretability: every prediction comes with a readable path of questions and answers. When a decision must be explained to a customer, auditor, or regulator, that path is the explanation.
How a Decision Tree Works
Training splits the dataset one question at a time. At each node, the algorithm picks the attribute that best separates the outcomes, scored by a criterion such as information gain, and branches the data accordingly. Splitting stops when the subsets are pure enough or a depth limit is reached.
Classification trees end in category labels (approve, reject); regression trees end in numeric values (a predicted price). The CART algorithm builds both.
The weakness: a tree grown deep memorizes its training data instead of learning general patterns. That tendency to overfitting is the main reason single trees are rarely deployed alone.
Decision Tree vs Random Forest
The practical difference: a decision tree is one readable model, while a random forest is hundreds of trees voting together. The forest wins on accuracy and stability; the single tree wins on interpretability. You trade one for the other.
| Criterion | Decision Tree | Random Forest |
|---|---|---|
| Structure | One tree | Hundreds of trees, each trained on a random data subset |
| Interpretability | Every prediction traceable as a rule path | A vote tally; not readable as rules |
| Overfitting risk | High when grown deep | Low; averaging cancels individual trees’ errors |
| Accuracy | Lower on complex data | Higher and more stable |
Example of a Decision Tree
A bank trains a tree on its past loans to screen applications. The root node asks whether the applicant’s credit score is above a threshold. If not, the next node asks for a co-signer; if so, it asks whether the requested amount exceeds a set share of annual income. Three questions in, the application lands on a leaf: approve or reject.
A rejected applicant asks why. The bank reads the path back: score below threshold, no co-signer. That answer is exactly what a black-box model cannot give, and it is why trees keep their place in regulated lending.
Related AI terms: Random Forest · Ensemble Learning · XGBoost · Interpretability
Did you like the Decision Tree gist?
Learn about 250+ need-to-know artificial intelligence terms in the AI Dictionary.
Mihail Sebastian — Writes about AI governance, regulation, and the technology behind them. Placeholder bio — replace with a real credential line. About