How does a Decision Tree make predictions?
Updated Feb 20, 2026
Short answer
A decision tree makes predictions by following a path of rules from the root node to a leaf node based on input features.
Deep explanation
When a new data point is given, the decision tree starts at the root node and checks the first condition (for example, “Is age > 30?”). Based on the answer, it moves down the corresponding branch. At each internal node, another condition is evaluated using a different feature. This process continues until the model reaches a leaf node, which contains the final prediction (such as a class label like “Approved” or a numeric value like “House price = 200K”).
The key idea is that each step reduces uncertainty by narrowing down possibilities. The tree essentially partitions the feature space into regions where each region corresponds to a specific output.
Real-world example
In medical diagnosis for diabetes:
- Is blood sugar > 140?
- Yes → Check BMI
- BMI > 30 → High risk
- BMI ≤ 30 → Moderate risk
- No → Low risk
- Yes → Check BMI
A patient’s test results are passed through this path to reach a diagnosis.
Common mistakes
- - Thinking the tree evaluates all conditions at once (it follows one path only).
- - Assuming predictions are probabilistic at every node (only leaves give final output).
- - Ignoring that small changes in data can change the path and prediction.
Follow-up questions
- What happens if a value is missing in a decision tree?
- How does a tree handle continuous vs categorical features?
- Why are decision trees easy to interpret?