gluonts.model package#

class gluonts.model.Estimator(lead_time: int = 0, **kwargs)[source]#

Bases: object

An abstract class representing a trainable model.

The underlying model is trained by calling the train method with a training Dataset, producing a Predictor object.

classmethod derive_auto_fields(train_iter)[source]#
classmethod from_hyperparameters(**hyperparameters)[source]#
classmethod from_inputs(train_iter, **params)[source]#
lead_time: int#
prediction_length: int#
train(training_data: Dataset, validation_data: Optional[Dataset] = None) → Predictor[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

class gluonts.model.Forecast[source]#

Bases: object

Abstract class representing predictions.

copy_aggregate(agg_fun: Callable)[source]#

Return a new Forecast object with a time series aggregated over the dimension axis.

Parameters

agg_fun – Aggregation function that defines the aggregation operation (typically mean or sum).

copy_dim(dim: int)[source]#

Return a new Forecast object with only the selected sub-dimension.

Parameters

dim – The returned forecast object will only represent this dimension.

dim() → int[source]#

Return the dimensionality of the forecast object.

property freq#
property index: PeriodIndex#
info: Optional[Dict]#
item_id: Optional[str]#
property mean: ndarray#
property median: ndarray#
plot(*, intervals=(0.5, 0.9), ax=None, color=None, name=None, show_label=False)[source]#

Plot median forecast and prediction intervals using matplotlib.

By default the 0.5 and 0.9 prediction intervals are plotted. Other intervals can be choosen by setting intervals.

This plots to the current axes object (via plt.gca()), or to ax if provided. Similarly, the color is using matplotlibs internal color cycle, if no explicit color is set.

One can set name to use it as the label for the median forecast. Intervals are not labeled, unless show_label is set to True.

prediction_length: int#
quantile(q: Union[float, str]) → ndarray[source]#

Compute a quantile from the predicted distribution.

Parameters

q – Quantile to compute.

Returns

Value of the quantile across the prediction range.

Return type

numpy.ndarray

quantile_ts(q: Union[float, str]) → Series[source]#
start_date: Period#
class gluonts.model.IncrementallyTrainable(*args, **kwargs)[source]#

Bases: Protocol

train_from(predictor: Predictor, training_data: Dataset, validation_data: Optional[Dataset] = None) → Predictor[source]#

Experimental: this feature may change in future versions. Train the estimator, starting from a previously trained predictor, on the given data.

Parameters

  • predictor – A previously trained model, from which to initialize the estimator training.

  • 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

class gluonts.model.Input(shape: Tuple[int, ...], dtype: Any, required: bool = True)[source]#

Bases: object

dtype: Any#
required: bool = True#
shape: Tuple[int, ...]#
class gluonts.model.InputSpec(data: Dict[str, gluonts.model.inputs.Input], zeros_fn: Callable)[source]#

Bases: UserDict

data: Dict[str, Input]#
property dtypes: Dict[str, Type]#
property shapes: Dict[str, Tuple[int, ...]]#
zeros()[source]#
zeros_fn: Callable#
class gluonts.model.Predictor(prediction_length: int, lead_time: int = 0)[source]#

Bases: object

Abstract class representing predictor objects. :param prediction_length: Prediction horizon.

classmethod derive_auto_fields(train_iter)[source]#
classmethod deserialize(path: Path, **kwargs) → Predictor[source]#

Load a serialized predictor from the given path.

Parameters

  • path – Path to the serialized files predictor.

  • **kwargs – Optional context/device parameter to be used with the predictor. If nothing is passed will use the GPU if available and CPU otherwise.

classmethod from_hyperparameters(**hyperparameters)[source]#
classmethod from_inputs(train_iter, **params)[source]#
predict(dataset: Dataset, **kwargs) → Iterator[Forecast][source]#

Compute forecasts for the time series in the provided dataset. This method is not implemented in this abstract class; please use one of the subclasses. :param dataset: The dataset containing the time series to predict.

Returns

Iterator over the forecasts, in the same order as the dataset iterable was provided.

Return type

Iterator[Forecast]

serialize(path: Path) → None[source]#
class gluonts.model.QuantileForecast(forecast_arrays: ndarray, start_date: Period, forecast_keys: List[str], item_id: Optional[str] = None, info: Optional[Dict] = None)[source]#

Bases: Forecast

A Forecast that contains arrays (i.e. time series) for quantiles and mean.

Parameters

  • forecast_arrays – An array of forecasts

  • start_date (pandas._libs.tslibs.period.Period) – start of the forecast

  • forecast_keys – A list of quantiles of the form ‘0.1’, ‘0.9’, etc., and potentially ‘mean’. Each entry corresponds to one array in forecast_arrays.

  • item_id (Optional[str]) – Identifier of the item being forecasted.

  • info (Optional[Dict]) – Additional information that the forecaster may provide e.g. estimated parameters, number of iterations ran etc.

copy_dim(dim: int) → QuantileForecast[source]#

Return a new Forecast object with only the selected sub-dimension.

Parameters

dim – The returned forecast object will only represent this dimension.

dim() → int[source]#

Return the dimensionality of the forecast object.

property mean: ndarray#

Forecast mean.

quantile(inference_quantile: Union[float, str]) → ndarray[source]#

Compute a quantile from the predicted distribution.

Parameters

q – Quantile to compute.

Returns

Value of the quantile across the prediction range.

Return type

numpy.ndarray

class gluonts.model.SampleForecast(samples: ndarray, start_date: Period, item_id: Optional[str] = None, info: Optional[Dict] = None)[source]#

Bases: Forecast

A Forecast object, where the predicted distribution is represented internally as samples.

Parameters

  • samples – Array of size (num_samples, prediction_length) (1D case) or (num_samples, prediction_length, target_dim) (multivariate case)

  • start_date (pandas._libs.tslibs.period.Period) – Start of the forecast.

  • item_id (Optional[str]) – Identifier of the item being forecasted.

  • info (Optional[Dict]) – Additional information that the forecaster may provide e.g. estimated parameters, number of iterations ran etc.

copy_aggregate(agg_fun: Callable) → SampleForecast[source]#

Return a new Forecast object with a time series aggregated over the dimension axis.

Parameters

agg_fun – Aggregation function that defines the aggregation operation (typically mean or sum).

copy_dim(dim: int) → SampleForecast[source]#

Return a new Forecast object with only the selected sub-dimension.

Parameters

dim – The returned forecast object will only represent this dimension.

dim() → int[source]#

Return the dimensionality of the forecast object.

property mean: ndarray#

Forecast mean.

property mean_ts: Series#

Forecast mean, as a pandas.Series object.

property num_samples#

The number of samples representing the forecast.

property prediction_length#

Time length of the forecast.

quantile(q: Union[float, str]) → ndarray[source]#

Compute a quantile from the predicted distribution.

Parameters

q – Quantile to compute.

Returns

Value of the quantile across the prediction range.

Return type

numpy.ndarray

to_quantile_forecast(quantiles: List[str]) → QuantileForecast[source]#
gluonts.model.evaluate_forecasts(forecasts: Iterable[Forecast], *, test_data: TestData, metrics, axis: Optional[Union[int, tuple]] = None, batch_size: int = 100, mask_invalid_label: bool = True, allow_nan_forecast: bool = False, seasonality: Optional[int] = None) → DataFrame[source]#

Evaluate forecasts by comparing them with test_data, according to metrics.

Note

This feature is experimental and may be subject to changes.

The optional axis arguments controls aggregation of the metrics: - None (default) aggregates across all dimensions - 0 aggregates across the dataset - 1 aggregates across the first data dimension (time, in the univariate setting) - 2 aggregates across the second data dimension (time, in the multivariate setting)

Return results as a Pandas DataFrame.

gluonts.model.evaluate_model(model: Predictor, *, test_data: TestData, metrics, axis: Optional[Union[int, tuple]] = None, batch_size: int = 100, mask_invalid_label: bool = True, allow_nan_forecast: bool = False, seasonality: Optional[int] = None) → DataFrame[source]#

Evaluate model when applied to test_data, according to metrics.

Note

This feature is experimental and may be subject to changes.

The optional axis arguments controls aggregation of the metrics: - None (default) aggregates across all dimensions - 0 aggregates across the dataset - 1 aggregates across the first data dimension (time, in the univariate setting) - 2 aggregates across the second data dimension (time, in the multivariate setting)

Return results as a Pandas DataFrame.

Submodules#