Home▸AI & Machine Learning▸Lag Features, Rolling Windows & Cyclical Encoding

📈 Lag Features, Rolling Windows & Cyclical Encoding

A 3D lab that turns a raw time-series signal into lag features, rolling-window statistics, and sine/cosine cyclical encodings — the inputs tree models actually learn from.

AI & Machine Learning3DAdvanced60 FPS

A raw time-series signal slides past a scoring cursor while orange lag markers, a translucent rolling window, and a rotating sine/cosine wheel show exactly how each engineered feature is derived before it ever reaches a model.

🔬 What It Demonstrates

Lag(1)…lag(N) bars copy past values into the current row, a rolling window computes trailing mean and ±1σ volatility, and a wheel decomposes cyclical time (hour/weekday/month) into sin and cos columns so periodic boundaries don't create false distance.

🎮 How to Use

Adjust lag steps and rolling window size, pick a cyclical period, and watch the live stats update as the series plays. Pause anytime to inspect a single row, or generate a fresh synthetic signal.

💡 Did You Know?

Computing lag or rolling features from data that includes future rows (instead of strictly the past) is one of the most common causes of time-series data leakage — it inflates validation scores that then collapse in production.

⚙ Under the hood

A 3D lab that turns a raw time-series signal into lag features, rolling-window statistics, and sine/cosine cyclical encodings — the inputs tree models actually learn from.

machine learningtime series analysisfeature engineeringdata modelingalgorithmsstatistical analysiscyclical encodingThree.js

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