Home▸AI & Machine Learning▸LSTM Airline Passenger Forecasting Lab

📈 LSTM Airline Passenger Forecasting Lab

An interactive 3D visualization of an LSTM reading a monthly airline-passenger time series: watch the lookback window slide, the hidden state react, and the forecast track (or miss) the real seasonal curve.

AI & Machine Learning3DAdvanced60 FPS

A 3D bar chart of monthly airline passenger totals with a sliding lookback window, an animated hidden-state cluster, and a live LSTM-style prediction tube that tightens or drifts as you change the network's memory and capacity.

🔬 What It Demonstrates

Prediction accuracy depends on both lookback window length and hidden-state size: short windows and small hidden layers under-fit the seasonal pattern, while wider context and more capacity let the blue prediction tube hug the grey actual bars.

🎮 How to Use

Adjust lookback window and hidden units to watch the mean absolute error change, then enable the 12-month forecast to see the amber uncertainty cone widen the further the model projects beyond real data.

💡 Did You Know?

The AirPassengers dataset (1949-1960) is one of the most-used toy datasets in time-series forecasting — it has trend, multiplicative seasonality, and nonstationary variance, all in just 144 points.

⚙ Under the hood

lstmtime seriesforecastingneural networksmachine learningairline passengersThree.js

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