Skip to content

BaseStatsForecaster

yohou_nixtla.stats.BaseStatsForecaster

Bases: BaseNixtlaForecaster

Base class for statsforecast model wrappers in yohou.

Wraps a statsforecast model class using the BaseNixtlaForecaster pattern, making it fully compatible with yohou's forecaster API (fit / predict / observe / clone / get_params).

Subclasses only need to set _estimator_default_class to a specific statsforecast model class (e.g., AutoARIMA).

Parameters

Name Type Description Default
model type or None

The statsforecast model class to wrap. Must be a subclass of statsforecast.models._TS. If not provided, _estimator_default_class is used.

 None
freq str or None

Frequency string (pandas offset alias). If None, auto-inferred from the data at fit time.

 None
**params dict

Parameters forwarded to the statsforecast model constructor.

 {}

Attributes

Name Type Description
nixtla_forecaster_ StatsForecast

The fitted Nixtla StatsForecast orchestrator (internal).
freq_ str

The inferred or provided frequency string.
instance_ _TS

The constructed statsforecast model instance (from BaseClassWrapper).
y_columns_ list of str

Original target column names from the training data.

See Also

yohou_nixtla.stats.AutoARIMAForecaster : AutoARIMA wrapper. yohou_nixtla.stats.NaiveForecaster : Naive baseline wrapper.

Source Code

View on GitHub

Show/Hide source 
 20
 21
 22
 23
 24
 25
 26
 27
 28
 29
 30
 31
 32
 33
 34
 35
 36
 37
 38
 39
 40
 41
 42
 43
 44
 45
 46
 47
 48
 49
 50
 51
 52
 53
 54
 55
 56
 57
 58
 59
 60
 61
 62
 63
 64
 65
 66
 67
 68
 69
 70
 71
 72
 73
 74
 75
 76
 77
 78
 79
 80
 81
 82
 83
 84
 85
 86
 87
 88
 89
 90
 91
 92
 93
 94
 95
 96
 97
 98
 99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
class BaseStatsForecaster(BaseNixtlaForecaster):
    """Base class for statsforecast model wrappers in yohou.

    Wraps a statsforecast model class using the ``BaseNixtlaForecaster``
    pattern, making it fully compatible with yohou's forecaster API
    (``fit`` / ``predict`` / ``observe`` / ``clone`` / ``get_params``).

    Subclasses only need to set ``_estimator_default_class`` to a specific
    statsforecast model class (e.g., ``AutoARIMA``).

    Parameters
    ----------
    model : type or None, default=None
        The statsforecast model class to wrap. Must be a subclass of
        ``statsforecast.models._TS``. If not provided,
        ``_estimator_default_class`` is used.
    freq : str or None, default=None
        Frequency string (pandas offset alias). If None, auto-inferred
        from the data at fit time.
    **params : dict
        Parameters forwarded to the statsforecast model constructor.

    Attributes
    ----------
    nixtla_forecaster_ : StatsForecast
        The fitted Nixtla StatsForecast orchestrator (internal).
    freq_ : str
        The inferred or provided frequency string.
    instance_ : _TS
        The constructed statsforecast model instance (from ``BaseClassWrapper``).
    y_columns_ : list of str
        Original target column names from the training data.

    See Also
    --------
    yohou_nixtla.stats.AutoARIMAForecaster : AutoARIMA wrapper.
    yohou_nixtla.stats.NaiveForecaster : Naive baseline wrapper.

    """

    _estimator_name = "model"
    _estimator_base_class = _TS
    _estimator_default_class: type | None = None

    @_fit_context(prefer_skip_nested_validation=True)
    def fit(
        self,
        y: pl.DataFrame,
        X_actual: pl.DataFrame | None = None,
        forecasting_horizon: int = 1,
        X_future: pl.DataFrame | None = None,
        X_forecast: pl.DataFrame | None = None,
        **params,
    ) -> BaseStatsForecaster:
        """Fit the statsforecast model to the training data.

        Parameters
        ----------
        y : pl.DataFrame
            Target time series with ``time`` column.
        X_actual : pl.DataFrame or None, default=None
            Actual observation features with ``time`` column.
        forecasting_horizon : int, default=1
            Number of steps to forecast.
        X_future : pl.DataFrame or None, default=None
            Known future features with ``time`` column.
        X_forecast : pl.DataFrame or None, default=None
            Not supported. Raises ``ValueError`` if provided.
        **params : dict
            Additional metadata routing parameters.

        Returns
        -------
        self
            Fitted forecaster.

        """
        return super().fit(
            y=y,
            X_actual=X_actual,
            forecasting_horizon=forecasting_horizon,
            X_future=X_future,
            X_forecast=X_forecast,
            **params,
        )

    def _fit_backend(self, nixtla_df: Any, forecasting_horizon: int) -> None:
        """Create and fit the StatsForecast orchestrator.

        Parameters
        ----------
        nixtla_df : pd.DataFrame
            Training data in Nixtla long format.
        forecasting_horizon : int
            Number of steps to forecast.

        """
        sf = StatsForecast(
            models=[self.instance_],
            freq=self.freq_,
        )
        sf.fit(df=nixtla_df)
        self.nixtla_forecaster_ = sf

    def _predict_backend(self, forecasting_horizon: int, X_future: Any = None) -> Any:
        """Generate raw predictions from the StatsForecast orchestrator.

        Parameters
        ----------
        forecasting_horizon : int
            Number of steps to forecast.
        X_future : pd.DataFrame or None, default=None
            Known future features in Nixtla long format, passed as
            ``X_df`` to ``StatsForecast.predict()``.

        Returns
        -------
        pd.DataFrame
            Raw statsforecast predictions.

        """
        check_is_fitted(self, ["nixtla_forecaster_"])
        return self.nixtla_forecaster_.predict(h=forecasting_horizon, X_df=X_future)

Methods

fit(y, X_actual=None, forecasting_horizon=1, X_future=None, X_forecast=None, **params)

Fit the statsforecast model to the training data.

Parameters
Name Type Description Default
y DataFrame

Target time series with time column.

 required
X_actual DataFrame or None

Actual observation features with time column.

 None
forecasting_horizon int

Number of steps to forecast.

 1
X_future DataFrame or None

Known future features with time column.

 None
X_forecast DataFrame or None

Not supported. Raises ValueError if provided.

 None
**params dict

Additional metadata routing parameters.

 {}
Returns
Type Description
self

Fitted forecaster.
Source Code

View on GitHub

Show/Hide source 
 64
 65
 66
 67
 68
 69
 70
 71
 72
 73
 74
 75
 76
 77
 78
 79
 80
 81
 82
 83
 84
 85
 86
 87
 88
 89
 90
 91
 92
 93
 94
 95
 96
 97
 98
 99
100
101
102
103
104
@_fit_context(prefer_skip_nested_validation=True)
def fit(
    self,
    y: pl.DataFrame,
    X_actual: pl.DataFrame | None = None,
    forecasting_horizon: int = 1,
    X_future: pl.DataFrame | None = None,
    X_forecast: pl.DataFrame | None = None,
    **params,
) -> BaseStatsForecaster:
    """Fit the statsforecast model to the training data.

    Parameters
    ----------
    y : pl.DataFrame
        Target time series with ``time`` column.
    X_actual : pl.DataFrame or None, default=None
        Actual observation features with ``time`` column.
    forecasting_horizon : int, default=1
        Number of steps to forecast.
    X_future : pl.DataFrame or None, default=None
        Known future features with ``time`` column.
    X_forecast : pl.DataFrame or None, default=None
        Not supported. Raises ``ValueError`` if provided.
    **params : dict
        Additional metadata routing parameters.

    Returns
    -------
    self
        Fitted forecaster.

    """
    return super().fit(
        y=y,
        X_actual=X_actual,
        forecasting_horizon=forecasting_horizon,
        X_future=X_future,
        X_forecast=X_forecast,
        **params,
    )