gluonts.nursery.autogluon_tabular.estimator module

class gluonts.nursery.autogluon_tabular.estimator.TabularEstimator(freq: str, prediction_length: int, lag_indices: Optional[List[int]] = None, time_features: Optional[List[gluonts.time_feature._base.TimeFeature]] = None, scaling: Callable[[pandas.core.series.Series], Tuple[pandas.core.series.Series, float]] = <function mean_abs_scaling>, batch_size: Optional[int] = 32, disable_auto_regression: bool = False, **kwargs)[source]

Bases: gluonts.model.estimator.Estimator

An estimator that trains an Autogluon Tabular model for time series forecasting.

Additional keyword arguments to the constructor, other than the ones documented below, will be passed on to Autogluon Tabular’s fit method used for training the model.

Parameters
  • freq – Frequency of the data to handle

  • prediction_length – Prediction length

  • lag_indices – List of indices of the lagged observations to use as features. If None, this will be set automatically based on the frequency.

  • time_features – List of time features to be used. If None, this will be set automatically based on the frequency.

  • scaling – Function to be used to scale time series. This should take a pd.Series object as input, and return a scaled pd.Series and the scale (float). By default, this divides a series by the mean of its absolute value.

  • batch_size – Batch size of the resulting predictor; this is just used at prediction time, and does not affect training in any way.

  • disable_auto_regression – Weather to forecefully disable auto-regression in the model. If True, this will remove any lag index which is smaller than prediction_length. This will make predictions more efficient, but may impact their accuracy.

freq = None
lead_time = None
prediction_length = None
train(training_data: Iterable[Dict[str, Any]], eval_metric='mean_absolute_error') → gluonts.nursery.autogluon_tabular.predictor.TabularPredictor[source]

Train the estimator on the given data.

Parameters
  • training_data – Dataset to train the model on.

  • validation_data – Dataset to validate the model on during training.

Returns

The predictor containing the trained model.

Return type

Predictor