gluonts.model.trivial.mean module#
- class gluonts.model.trivial.mean.MeanEstimator(prediction_length: PositiveInt, num_samples: PositiveInt)[source]#
Bases:
Estimator
An Estimator that computes the mean targets in the training data, in the trailing prediction_length observations, and produces a ConstantPredictor that always predicts such mean value.
- Parameters:
prediction_length (int) – Prediction horizon.
num_samples – Number of samples to include in the forecasts. Not that the samples produced by this predictor will all be identical.
- train(training_data: Dataset, validation_dataset: Optional[Dataset] = None) ConstantPredictor [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:
- class gluonts.model.trivial.mean.MeanPredictor(prediction_length: int, num_samples: int = 100, context_length: Optional[int] = None)[source]#
Bases:
RepresentablePredictor
A
Predictor
that predicts the samples based on the mean of the last context_length elements of the input target.- Parameters:
context_length – Length of the target context used to condition the predictions.
prediction_length – Length of the prediction horizon.
num_samples – Number of samples to use to construct
SampleForecast
objects for every prediction.
- predict_item(item: Dict[str, Any]) SampleForecast [source]#
- class gluonts.model.trivial.mean.MovingAveragePredictor(prediction_length: int, context_length: Optional[int] = None)[source]#
Bases:
RepresentablePredictor
A
Predictor
that predicts the moving average based on the last context_length elements of the input target.If prediction_length = 1, the output is the moving average based on the last context_length elements of the input target.
If prediction_length > 1, the output is the moving average based on the last context_length elements of the input target, where previously calculated moving averages are appended at the end of the inputtarget. Hence, for prediction_length larger than context_length, there will be cases where the moving average is calculated on top of previous moving averages.
- Parameters:
context_length – Length of the target context used to condition the predictions.
prediction_length – Length of the prediction horizon.
- predict_item(item: Dict[str, Any]) SampleForecast [source]#