Knowledge MaxxingFeature scaling
Feature scaling transforms numeric features onto a common range or distribution, so features measured in different units contribute fairly to a model.
Prerequisite: Feature
What matters
Many algorithms that rely on distance or magnitude, like k-means or gradient descent, can be dominated by a feature with naturally larger numeric values if scaling is skipped.
Example
Without scaling, a feature measured in dollars (thousands) can overwhelm a feature measured as a 0-to-1 ratio in a distance-based algorithm, even if the ratio matters just as much.
Common misconception
Feature scaling is not needed for every algorithm; tree-based models like decision trees and random forests are largely unaffected by the scale of numeric features.
Recall question
Why can an unscaled feature with large values dominate a distance-based algorithm?