What is feature scaling in tree-based models and is it necessary?
Updated May 16, 2026
Short answer
Tree-based models do not require feature scaling because they split based on thresholds.
Deep explanation
Decision trees, random forests, and gradient boosting models split data based on feature thresholds rather than distances. Therefore, scaling does not affect their performance, unlike linear or neural models.
Real-world example
Used in credit risk scoring where tree models dominate.
Common mistakes
- Applying unnecessary scaling to tree-based pipelines.
Follow-up questions
- Why are trees scale-invariant?
- When might scaling still matter?