Home▸Data Science▸Categorical Encoding Lab

🏷️ Categorical Encoding Lab

Switch a categorical feature between one-hot, label, frequency and target encoding and watch what the model actually sees change live — including the overfitting risk of target encoding on small categories.

Data Science3DModerate60 FPS

Details

CategoryData Science

RenderingThree.js · WebGL

DifficultyModerate

Target FPS60

ControlsCardinality slider (3-10 categories) sets how many distinct category values exist; Samples-per-category slider (4-120) controls how much data backs each category's target-encoding estimate — fewer samples widens the translucent uncertainty band around the target-encoded bar; One-Hot / Label / Frequency / Target buttons switch the live encoding method, instantly re-deriving every bar height and the dimensionality/order-correlation/noise readouts; 'New random data' regenerates the underlying categories and target values; drag to orbit the 3D scene.

Three.jsData ScienceFeature EngineeringCategorical EncodingOne-Hot EncodingTarget EncodingMachine Learning

Related simulations

Feature Engineering Transform GraphRaw feature nodes flow through transform nodes (scaling, encoding, interaction terms) into…

Anomaly Detection Point CloudNormal data points cluster tightly while injected anomalies drift outward — tune detection…

Data Science: A Physics PerspectiveExplore the core concepts of data science – analytics, modeling, and machine learning –… ⊞ All 48 in Data Science →

⚙ Under the hood

Switch a categorical feature between one-hot, label, frequency and target encoding and watch what the model actually sees change live — including the growing overfitting risk of target encoding as category sample size shrinks.

Three.jsData ScienceFeature EngineeringCategorical EncodingOne-Hot EncodingTarget EncodingMachine Learning

3D · Three.js / WebGL renderer · 60 FPS target · runs fully client-side, no install