Upload a CSV or Excel time series, pick a time grain and forecast duration — every model is trained and scored on a held-out validation window, and the winner is chosen automatically.
One row per timestamp. Needs a date/time column and at least one numeric column to forecast.
Each model is trained on the earlier data and scored on the held-out tail. Lower is better.
How each model tracked the actual held-out period.
Best model retrained on the full series. Shaded band = 90% forecast interval.