SBA Forecasting

SbaForecaster implements the Syntetos-Boylan approximation for intermittent demand. It starts from Croston-style demand and interval smoothing, then applies the standard bias adjustment to reduce Croston's upward bias on sparse series.

Interactive Example

SBA sparse-demand forecast

Runs sba against a bundled taxi route-demand sample.

Ready to run in this page.

Python Example

from cartoboost.forecasting import SbaForecaster

demand = [0, 0, 4, 0, 0, 7, 0, 3, 0, 0, 0, 5]

model = SbaForecaster(alpha=0.2)
model.fit(demand)
forecast = model.predict(6)

Use When

Use SBA when Croston is a reasonable sparse-demand baseline but you want the bias-adjusted level that usually forecasts below the unadjusted Croston value for the same alpha.

Fit Contract

Requirement	Detail
Target values	Finite and non-negative.
Zero values	Treated as true no-demand periods.
Non-zero demand	At least one non-zero observation is required.
Smoothing	alpha must be greater than 0 and at most 1.

Validation

SBA is still a fixed local method. Validate it against Croston, TSB, seasonal naive, and any broader selector under the same cutoff and horizon.

Report the zero fraction, demand-event count, MAE or RMSE, WAPE when defined, and errors by horizon. Use identical smoothing-selection rules for Croston and SBA.

Limitations

• SBA produces a flat horizon forecast and has no calendar or covariate model.
• Bias correction does not guarantee lower holdout error for every series.
• Results are unreliable when the history contains too few non-zero events.