Holt-Winters parameters

Level smoothing α: 0.30 Trend smoothing β: 0.10 Seasonal smoothing γ: 0.30

Occupancy generator

Noise level: 0.05 Trend volatility: 0.40 Simulation speed: 1.0× (sim-hours/s)

Live state

Sim time—

Actual occupancy—

1-step forecast—

Level (l)—

Trend (b)—

Seasonal (s, this hour)—

Rolling RMSE (cars)—

Rolling RMSE (%)—

Capacity—

This is a genuine additive Holt-Winters triple exponential smoothing model: level, trend and a 24-hour seasonal cycle are updated every simulated hour from a synthetic parking-lot occupancy stream (daily rush-hour seasonality + slow random-walk trend + noise). The rolling RMSE compares each one-step-ahead forecast — made before the actual value arrived — against what actually happened, so it is a genuine held-out accuracy score, not a fit statistic. Drag α/β/γ to see the forecast track (or fail to track) the live data.