CrostonClassic

chronax.croston_classic.CrostonClassic · inherits BaseForecaster

Croston's Classic method for intermittent demand time series.

Suitable for series with many zero values and occasional non-zero demand. Uses SES (α=0.1) to forecast both demand size and inter-demand intervals.

Key Features: - Handles sparse/intermittent data (many zeros) - Fixed smoothing parameter α=0.1 (Croston's original specification) - Decomposes series into demand size and demand intervals - Conformal prediction intervals supported

Note: Only conformal intervals are supported (no native parametric intervals).

__init__(self, alias='CrostonClassic', conformal_params=None)

Initializes the CrostonClassic model.

Parameter	Type	Default	Description
alias	str	"CrostonClassic"	Model name.
conformal_params	Optional[ConformalIntervals]	None	Configuration for conformal prediction intervals.

fit(self, y, X=None) -> Self

Fit Croston Classic model to time series.

Parameter	Type	Default	Description
y	jnp.ndarray	-	Time series of shape (t,).
X	Optional[jnp.ndarray]	None	Unused (no exogenous support).

Returns: Self (the fitted forecaster; sets self.model_).

predict(self, h, X=None, level=None) -> Dict[str, jnp.ndarray]

Generate forecasts using fitted model.

Parameter	Type	Default	Description
h	int	-	Forecast horizon.
X	Optional[jnp.ndarray]	None	Unused (no exogenous support).
level	Optional[List[int]]	None	Confidence levels for prediction intervals (0-100).

Returns: dict (Dictionary with 'mean' and optional interval keys ('lo-XX', 'hi-XX')).

predict_in_sample(self, level=None) -> Dict[str, jnp.ndarray]

Access fitted (in-sample) predictions.

Note: Native (parametric) fitted intervals are supported using residual std error.

Parameter	Type	Default	Description
level	Optional[List[int]]	None	Confidence levels for fitted intervals (0-100).

Returns: dict (Dictionary with 'fitted' and optional interval keys).

forecast(self, y, h, X=None, X_future=None, level=None, fitted=False) -> Dict[str, jnp.ndarray]

Memory-efficient forecast without storing model state.

Equivalent to fit().predict() but avoids object storage overhead. Useful for one-shot forecasting or cross-validation loops.

Parameter	Type	Default	Description
y	jnp.ndarray	-	Time series of shape (t,).
h	int	-	Forecast horizon.
X	Optional[jnp.ndarray]	None	Unused (no exogenous support).
X_future	Optional[jnp.ndarray]	None	Unused (no exogenous support).
level	Optional[List[int]]	None	Confidence levels for prediction intervals (0-100).
fitted	bool	False	Whether to return in-sample fitted values.

Returns: dict (Dictionary with 'mean', optional 'fitted', and interval keys).