Univariate Prophet

Bases: Prophetverse

Prophet forecaster implemented in Numpyro.

This forecaster uses a logistic trend and supports custom feature transformers for additional seasonality or holiday effects.

Parameters:

Name	Type	Description	Default
feature_transformer	sktime transformer	Transformer to generate additional features.	None
noise_scale	float	Scale parameter for observation noise. (default: 0.05)	0.05
trend	str	Type of trend, either "linear" or "logistic". (default: "logistic")	'logistic'
exogenous_effects	optional	List of exogenous effect objects.	None
default_effect	optional	Default effect for variables without a specified effect.	None
scale	optional	Scaling factor inferred from data.	None
rng_key	optional	A jax.random.PRNGKey instance, or None.	None

Source code in src/prophetverse/sktime/univariate.py

class Prophet(Prophetverse):
    """Prophet forecaster implemented in Numpyro.

    This forecaster uses a logistic trend and supports custom feature transformers
    for additional seasonality or holiday effects.

    Parameters
    ----------
    feature_transformer : sktime transformer, optional
        Transformer to generate additional features.
    noise_scale : float, optional
        Scale parameter for observation noise. (default: 0.05)
    trend : str, optional
        Type of trend, either "linear" or "logistic". (default: "logistic")
    exogenous_effects : optional
        List of exogenous effect objects.
    default_effect : optional
        Default effect for variables without a specified effect.
    scale : optional
        Scaling factor inferred from data.
    rng_key : optional
        A jax.random.PRNGKey instance, or None.
    """

    def __init__(
        self,
        feature_transformer=None,
        noise_scale=0.05,
        trend="logistic",
        exogenous_effects=None,
        default_effect=None,
        scale=None,
        rng_key=None,
        inference_engine=None,
    ):
        super().__init__(
            feature_transformer=feature_transformer,
            noise_scale=noise_scale,
            trend=trend,
            exogenous_effects=exogenous_effects,
            likelihood="normal",
            default_effect=default_effect,
            scale=scale,
            rng_key=rng_key,
            inference_engine=inference_engine,
        )