class AutoARIMA(ExtendedForecastingModel):
def __init__(self, *autoarima_args, **autoarima_kwargs):
""" Auto-ARIMA
This implementation is a thin wrapper around `pmdarima.txt AutoARIMA model
<https://alkaline-ml.com/pmdarima/modules/generated/pmdarima.arima.AutoARIMA.html>`_,
which provides functionality similar to R's `auto.arima
<https://www.rdocumentation.org/packages/forecast/versions/7.3/topics/auto.arima>`_.
This model supports the same parameters as the pmdarima.txt AutoARIMA model.
See `pmdarima.txt documentation
<https://alkaline-ml.com/pmdarima/modules/generated/pmdarima.arima.AutoARIMA.html>`_
for an extensive documentation and a list of supported parameters.
Parameters
----------
autoarima_args
Positional arguments for the pmdarima.txt AutoARIMA model
autoarima_kwargs
Keyword arguments for the pmdarima.txt AutoARIMA model
"""
super().__init__()
self.model = PmdAutoARIMA(*autoarima_args, **autoarima_kwargs)
self.trend = self.model.trend
def __str__(self):
return 'Auto-ARIMA'
def fit(self, series: TimeSeries, exog: Optional[TimeSeries] = None):
super().fit(series, exog)
series = self.training_series
self.model.fit(series.values(), X=exog.values() if exog else None)
def predict(self,
n: int,
exog: Optional[TimeSeries] = None,
num_samples: int = 1):
super().predict(n, exog, num_samples)
forecast = self.model.predict(n_periods=n,
X=exog.values() if exog else None)
return self._build_forecast_series(forecast)
@property
def min_train_series_length(self) -> int:
return 30
def _supports_range_index(self) -> bool:
raise_if(
self.trend and self.trend != "c",
"'trend' is not None. Range indexing is not supported in that case.",
logger)
return True
def create_model(self, **kwargs):
if self.kind == ModelT.holt_winters.name:
data = self.node.item
model = ExponentialSmoothing(endog=data, **kwargs)
elif self.kind == ModelT.auto_arima.name:
try:
from pmdarima import AutoARIMA
except ImportError: # pragma: no cover
logger.error(
"pmdarima not installed, so auto_arima won't work. Exiting."
"Install it with: pip install scikit-hts[auto_arima]")
return
model = AutoARIMA(**kwargs)
elif self.kind == ModelT.sarimax.name:
as_df = self.node.item
end = self.node.get_series()
if self.node.exogenous:
ex = as_df[self.node.exogenous]
else:
ex = None
model = SARIMAX(endog=end, exog=ex, **kwargs)
else:
raise
return model
def __init__(self, *autoarima_args, **autoarima_kwargs):
"""Auto-ARIMA
This implementation is a thin wrapper around `pmdarima AutoARIMA model
<https://alkaline-ml.com/pmdarima/modules/generated/pmdarima.arima.AutoARIMA.html>`_,
which provides functionality similar to R's `auto.arima
<https://www.rdocumentation.org/packages/forecast/versions/7.3/topics/auto.arima>`_.
This model supports the same parameters as the pmdarima AutoARIMA model.
See `pmdarima documentation
<https://alkaline-ml.com/pmdarima/modules/generated/pmdarima.arima.AutoARIMA.html>`_
for an extensive documentation and a list of supported parameters.
.. note::
For a faster and probabilistic version of AutoARIMA, checkout
the :class:`StatsForecastAutoARIMA` model.
Parameters
----------
autoarima_args
Positional arguments for the pmdarima.AutoARIMA model
autoarima_kwargs
Keyword arguments for the pmdarima.AutoARIMA model
"""
super().__init__()
self.model = PmdAutoARIMA(*autoarima_args, **autoarima_kwargs)
self.trend = self.model.trend
class AutoARIMA(UnivariateForecastingModel):
def __init__(self, *autoarima_args, **autoarima_kwargs):
""" Auto-ARIMA
This implementation is a thin wrapper around `pmdarima.txt AutoARIMA model
<https://alkaline-ml.com/pmdarima/modules/generated/pmdarima.arima.AutoARIMA.html>`_,
which provides functionality similar to R's `auto.arima
<https://www.rdocumentation.org/packages/forecast/versions/7.3/topics/auto.arima>`_.
This model supports the same parameters as the pmdarima.txt AutoARIMA model.
See `pmdarima.txt documentation
<https://alkaline-ml.com/pmdarima/modules/generated/pmdarima.arima.AutoARIMA.html>`_
for an extensive documentation and a list of supported parameters.
Parameters
----------
autoarima_args
Positional arguments for the pmdarima.txt AutoARIMA model
autoarima_kwargs
Keyword arguments for the pmdarima.txt AutoARIMA model
"""
super().__init__()
self.model = PmdAutoARIMA(*autoarima_args, **autoarima_kwargs)
def __str__(self):
return 'Auto-ARIMA'
def fit(self, series: TimeSeries):
super().fit(series)
series = self.training_series
self.model.fit(series.values())
def predict(self, n):
super().predict(n)
forecast = self.model.predict(n_periods=n)
return self._build_forecast_series(forecast)
@property
def min_train_series_length(self) -> int:
return 30
def create_model(self, **kwargs):
if self.kind == Model.holt_winters.name:
data = self.node.item
model = ExponentialSmoothing(endog=data, **kwargs)
elif self.kind == Model.auto_arima.name:
model = AutoARIMA(**kwargs)
elif self.kind == Model.sarimax.name:
as_df = self.node.item
end = self.node.get_series()
if self.node.exogenous:
ex = as_df[self.node.exogenous]
else:
ex = None
model = SARIMAX(endog=end, exog=ex, **kwargs)
else:
raise
return model
series_count=3,
series_size=365 * 3,
start_dt="2019-01-01",
days_period=1,
)
training_data = generated_data.df
group_key_columns = generated_data.key_columns
pipeline = Pipeline(
steps=[
(
"arima",
AutoARIMA(
max_order=14,
out_of_sample_size=90,
suppress_warnings=True,
error_action="ignore",
),
)
]
)
diff_analyzer = PmdarimaAnalyzer(
df=training_data,
group_key_columns=group_key_columns,
y_col="y",
datetime_col="ds",
)
ndiff = diff_analyzer.calculate_ndiffs(
alpha=0.05,
test="kpss",
class AutoARIMA(DualCovariatesForecastingModel):
def __init__(self, *autoarima_args, **autoarima_kwargs):
"""Auto-ARIMA
This implementation is a thin wrapper around `pmdarima AutoARIMA model
<https://alkaline-ml.com/pmdarima/modules/generated/pmdarima.arima.AutoARIMA.html>`_,
which provides functionality similar to R's `auto.arima
<https://www.rdocumentation.org/packages/forecast/versions/7.3/topics/auto.arima>`_.
This model supports the same parameters as the pmdarima AutoARIMA model.
See `pmdarima documentation
<https://alkaline-ml.com/pmdarima/modules/generated/pmdarima.arima.AutoARIMA.html>`_
for an extensive documentation and a list of supported parameters.
.. note::
For a faster and probabilistic version of AutoARIMA, checkout
the :class:`StatsForecastAutoARIMA` model.
Parameters
----------
autoarima_args
Positional arguments for the pmdarima.AutoARIMA model
autoarima_kwargs
Keyword arguments for the pmdarima.AutoARIMA model
"""
super().__init__()
self.model = PmdAutoARIMA(*autoarima_args, **autoarima_kwargs)
self.trend = self.model.trend
def __str__(self):
return "Auto-ARIMA"
def _fit(self,
series: TimeSeries,
future_covariates: Optional[TimeSeries] = None):
super()._fit(series, future_covariates)
series = self.training_series
self.model.fit(
series.values(),
X=future_covariates.values() if future_covariates else None)
return self
def _predict(
self,
n: int,
future_covariates: Optional[TimeSeries] = None,
num_samples: int = 1,
):
super()._predict(n, future_covariates, num_samples)
forecast = self.model.predict(
n_periods=n,
X=future_covariates.values() if future_covariates else None)
return self._build_forecast_series(forecast)
@property
def min_train_series_length(self) -> int:
return 10
def _supports_range_index(self) -> bool:
raise_if(
self.trend and self.trend != "c",
"'trend' is not None. Range indexing is not supported in that case.",
logger,
)
return True