Home▸AI & Machine Learning▸Feature Engineering and Data Preprocessing

🧪 Feature Engineering and Data Preprocessing

A 3D scatter of a non-linearly-separable dataset that you reshape live with scaling, outlier clipping and an engineered polynomial feature until a flat plane can separate the two classes.

AI & Machine Learning3DAdvanced60 FPS

A 3D scatter of two rings of data that cannot be separated by any flat line in their raw 2D form — reshape them live with scaling, outlier clipping and an engineered polynomial feature until a single plane cleanly separates the classes.

🔬 What It Demonstrates

Engineering a new feature (x² + y²) lifts the outer ring above the inner cluster, turning a non-linear problem into a linearly separable one — the same idea behind polynomial features and kernel tricks. Scaling and outlier handling reshape the same cloud the way a real preprocessing pipeline would.

🎮 How to Use

Drag the polynomial-feature slider to lift the outer ring, pick a scaling method, inject outliers and toggle clipping, then watch the live "linear separability" stat and the translucent separating plane respond.

💡 Did You Know?

Unscaled outliers can single-handedly dominate a min-max range, squashing 95% of "normal" values into a tiny sliver near zero — which is exactly why robust clipping or scaling is a standard first step in most preprocessing pipelines.

⚙ Under the hood

A 3D scatter of a non-linearly-separable dataset that you reshape live with scaling, outlier clipping and an engineered polynomial feature until a flat plane can separate the two classes.

machine learningdata preprocessingfeature engineeringdimensionality reductionpolynomial featuresclusteringclassificationThree.js

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