juniorFeature Engineering
What is feature scaling vs normalization difference?
Updated May 16, 2026
Short answer
Feature scaling standardizes values; normalization rescales them to a fixed range like 0–1.
Deep explanation
Scaling typically refers to standardization (mean 0, std 1), while normalization rescales values to a bounded range. Both aim to improve model convergence and prevent dominance of large-scale features.
Real-world example
Used in neural networks where gradient descent requires stable feature ranges.
Common mistakes
- Using scaling blindly without checking algorithm requirements.
Follow-up questions
- Which is better for neural networks?
- Do tree models need scaling?