Data & Analysis

Feature Scaling Playground

Different features live on different scales: Age in years, Salary in thousands. But algorithms like KNN and SVM don't know this. Watch what happens when you scale.

Raw

Min-Max (0–1)

Normalization (mean)

Standardized (z-score)

Distance comparison

Raw: No transformation applied; features remain on their original scales.

Click two points to compare distances across scaling methods

Raw (No Scaling)

Features retain their original units: Age in years, Salary in dollars. Distance-based algorithms will be dominated by whichever feature has the largest absolute range.

Pros

+Interpretable values
+No information loss
+Works fine for tree-based models

Cons

−Distance algorithms skewed by scale
−Gradient descent converges slowly
−Regularization applies unequally

Common algorithms

Decision TreesRandom ForestsNaive Bayes

Applied Formula

x_raw = x  (no transformation)
y_raw = y

No transformation applied; features remain on their original scales.

Distance Between Two Points

Click two points on the scatter plot to compare Euclidean distances across scaling methods.

d = √((x₁−x₂)² + (y₁−y₂)²)

When to Use Which

Raw: Tree-based models: splits don't care about scale

Min-Max (0–1): Neural networks, bounded inputs, image pixel values

Normalization (mean): Gradient descent: centered and bounded without std

Standardized (z-score): SVM, PCA, KNN, linear regression, regularized models

← Previous Guide Next: Outlier Detection: Spot the Odd One Out →