SARIMAX

SARIMAX extends SARIMA to include external regressors (exogenous variables) alongside the time series itself. These can be event flags, promotions, weather variables, or other external drivers of the target.

When to use:

• Forecasting where external variables (promotions, holidays, weather) influence the target
• Demand forecasting with marketing or economic drivers
• Causal time series modeling

Input:

• Trained model checkpoint — exported SARIMAX fit
• Preprocessing config — scaling settings
• Training tail — last N observations
• Exogenous data — future values of external regressors for the forecast horizon
• Steps — forecast horizon

Output: Forecasted values accounting for external variables

Model Settings (set during training, used at inference)

AR Order (p), Differencing (d), MA Order (q) — non-seasonal ARIMA orders Seasonal P, D, Q, m — seasonal orders and period Trend (default: null) Deterministic trend term. n = no trend, c = constant, t = linear trend.

Inference Settings

No dedicated inference-time settings. Exogenous values for the forecast horizon must be provided at inference.