What Is Feature Scaling?

Feature Scaling Explainer

Feature scaling changes numeric variables to comparable ranges or distributions. Without it, a feature measured in thousands can dominate distance calculations or optimization steps even when its scale does not make it more informative.

Standardization subtracts a training-set mean and divides by a training-set standard deviation. Min-max scaling maps observed training values into a selected range. Robust approaches use statistics less sensitive to extreme values.

Scaling often matters for nearest-neighbor methods, support vector machines, principal-component methods, and gradient-based models. Tree-based models generally split values by order and are usually less affected by simple differences in units.

Scaler parameters must be learned from training data and then reused unchanged for validation, testing, and production. Outliers, shifting distributions, and bounded variables still require domain judgment rather than an automatic choice of transformer.

Feature scaling transforms numeric variables so differences in units or magnitude do not dominate a model unintentionally.

Key terms include feature scale, standardization, normalization, min-max scaling, robust scaling, mean, standard deviation and data leakage.

Standardization centers values and expresses them in units of training-set standard deviation.

Min-max and robust scaling serve different distributional needs and should not be treated as interchangeable defaults.

Fitting a scaler on all data allows held-out values to influence model development.

Production must reuse the saved training transformation and monitor whether new values have shifted beyond its assumptions.

Explore more "Explainers"

Discover additional explainers across politics, science, business, technology, and other fields. Each explainer breaks down a complex idea into clear, everyday language—helping you better understand how major concepts, systems, and debates shape the world around us.