Skip to content

AutoARIMAForecaster

yohou_nixtla.stats.AutoARIMAForecaster

Bases: BaseStatsForecaster

AutoARIMA forecaster via statsforecast.

Automatically selects the best ARIMA model using the Hyndman-Khandakar algorithm.

Parameters

Name Type Description Default
season_length int

Length of the seasonal period.

 1
d int or None

Order of first differencing. Auto-detected if None.

 None
D int or None

Order of seasonal differencing. Auto-detected if None.

 None
max_p int

Maximum non-seasonal AR order.

 5
max_q int

Maximum non-seasonal MA order.

 5
freq str or None

Frequency string. Auto-inferred from data if None.

 None
feature_transformer BaseTransformer or None

Transformer applied to exogenous features before fitting/predicting.

 None
target_transformer BaseTransformer or None

Transformer applied to the target before fitting. Inverse-transformed after predicting to return forecasts in the original scale.

 None
target_as_feature ('transformed', 'raw')

Whether to include target values as additional features.

 "transformed"
**params dict

Additional parameters forwarded to statsforecast.models.AutoARIMA.

 {}

Attributes

Name Type Description
nixtla_forecaster_ StatsForecast

The fitted Nixtla orchestrator.
instance_ AutoARIMA

The constructed AutoARIMA model instance.

See Also

ARIMAForecaster : ARIMA with manually specified orders. AutoETSForecaster : Automatic ETS model selection.

Examples

>>> from yohou_nixtla.stats import AutoARIMAForecaster
>>> forecaster = AutoARIMAForecaster(season_length=12)
>>> forecaster
AutoARIMAForecaster(...)

Source Code

View on GitHub

Show/Hide source 
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
class AutoARIMAForecaster(BaseStatsForecaster):
    """AutoARIMA forecaster via statsforecast.

    Automatically selects the best ARIMA model using the Hyndman-Khandakar
    algorithm.

    Parameters
    ----------
    season_length : int, default=1
        Length of the seasonal period.
    d : int or None, default=None
        Order of first differencing. Auto-detected if None.
    D : int or None, default=None
        Order of seasonal differencing. Auto-detected if None.
    max_p : int, default=5
        Maximum non-seasonal AR order.
    max_q : int, default=5
        Maximum non-seasonal MA order.
    freq : str or None, default=None
        Frequency string. Auto-inferred from data if None.
    feature_transformer : BaseTransformer or None, default=None
        Transformer applied to exogenous features before fitting/predicting.
    target_transformer : BaseTransformer or None, default=None
        Transformer applied to the target before fitting. Inverse-transformed
        after predicting to return forecasts in the original scale.
    target_as_feature : {"transformed", "raw"} or None, default=None
        Whether to include target values as additional features.
    **params : dict
        Additional parameters forwarded to ``statsforecast.models.AutoARIMA``.

    Attributes
    ----------
    nixtla_forecaster_ : StatsForecast
        The fitted Nixtla orchestrator.
    instance_ : AutoARIMA
        The constructed AutoARIMA model instance.

    See Also
    --------
    ARIMAForecaster : ARIMA with manually specified orders.
    AutoETSForecaster : Automatic ETS model selection.

    Examples
    --------
    >>> from yohou_nixtla.stats import AutoARIMAForecaster
    >>> forecaster = AutoARIMAForecaster(season_length=12)
    >>> forecaster  # doctest: +ELLIPSIS
    AutoARIMAForecaster(...)

    """

    _estimator_default_class = AutoARIMA

    _tags = {"requires_exogenous": True, "supports_exogenous": True}

Tutorials

The following example notebooks use this component:

  • How to Compare Forecasters

    Fit multiple statistical forecasters on the same dataset and evaluate their accuracy with MAE.

    View · Open in marimo