ThetaEstimationMethod

theta_model.ThetaEstimationMethod · inherits StrEnum

StrEnum avoids Pydantic 2.12+ ""class not fully defined"" with Literal under PEP 563.

Member	Value
least_squares	"least_squares"
correlation_optimal	"correlation_optimal"

ThetaHyperparams

theta_model.ThetaHyperparams · inherits PydanticBaseModel

Configuration structure for Theta Model hyperparameters.

Attributes

Attribute	Type	Default	Description
sp	int	-	Seasonal period
theta_method	ThetaEstimationMethod	-	Method for theta estimation
use_reduced_rank	bool	False	Whether to use cointegration/reduced rank

ThetaModel

theta_model.ThetaModel · inherits BaseModel

Theta model for univariate, multivariate, and covariate (stacked exogenous) benchmarks.

Multivariate tasks (multiple joint targets, no exogenous stack) fit one independent ThetaForecaster per target column. Covariate tasks (optional past_covariates / x_context) are concatenated during a joint Θ-line pipeline.

__init__(self, params: Dict[str, Any], settings: Dict[str, Any])

Initializes the ThetaModel, setting up internal state and parsing hyperparameters using ThetaHyperparams.

Parameter	Type	Default	Description
params	Dict[str, Any]	-	(undocumented)
settings	Dict[str, Any]	-	(undocumented)

train(self, y_context: np.ndarray, y_target: np.ndarray, timestamps_context: np.ndarray, timestamps_target: np.ndarray, x_context: Optional[np.ndarray] = None, x_target: Optional[np.ndarray] = None, past_covariates: Optional[np.ndarray] = None, **kwargs: dict) -> "ThetaModel"

Trains the Theta model on the provided context data. Handles independent multivariate fitting or joint Θ-matrix estimation based on the presence of covariates and data size.

Parameter	Type	Default	Description
y_context	np.ndarray	-	(undocumented)
y_target	np.ndarray	-	(undocumented)
timestamps_context	np.ndarray	-	(undocumented)
timestamps_target	np.ndarray	-	(undocumented)
x_context	Optional[np.ndarray]	None	(undocumented)
x_target	Optional[np.ndarray]	None	(undocumented)
past_covariates	Optional[np.ndarray]	None	(undocumented)
**kwargs	dict	-	(undocumented)

Returns: Self (The fitted ThetaModel instance).

predict(self, y_context: np.ndarray, timestamps_context: np.ndarray, timestamps_target: np.ndarray, x_context: Optional[np.ndarray] = None, x_target: Optional[np.ndarray] = None, past_covariates: Optional[np.ndarray] = None, **kwargs: dict) -> np.ndarray

Generates forecasts for the target horizon based on the fitted model and context data.

Parameter	Type	Default	Description
y_context	np.ndarray	-	(undocumented)
timestamps_context	np.ndarray	-	(undocumented)
timestamps_target	np.ndarray	-	(undocumented)
x_context	Optional[np.ndarray]	None	(undocumented)
x_target	Optional[np.ndarray]	None	(undocumented)
past_covariates	Optional[np.ndarray]	None	(undocumented)
**kwargs	dict	-	(undocumented)

Returns: np.ndarray (The forecast mean, shape (forecast_horizon, num_original_targets)).

Raises:

Exception	Description
ValueError	If the model is not trained yet.
ValueError	If past_covariates requirements are violated based on training configuration.