Next Forecast
A ridge regression baseline predicts the next 24 hours of simulated electricity demand. It uses daily and weekly lags, a trailing mean and calendar features. The exported checkpoint contains the fitted coefficients.
| Method | MAE | Evening MAE (17:00-21:59) |
|---|---|---|
| Next Forecast | 3.747 | 3.899 |
| Previous week, same hour | 3.806 | 3.835 |
The first 112 days train the model. The final 28 days supply 672 hourly targets. Each forecast begins at midnight and uses observations available at that time. The model improves overall MAE on this synthetic series. The weekly baseline has lower evening error.
Install NumPy and run python forecast.py to regenerate the checkpoint and evaluation. The Space displays the held-out forecasts. The reported scores apply to the simulated series in this repository.
Inference Providers NEW
This model isn't deployed by any Inference Provider. π Ask for provider support