LTSFTransformerForecaster

class LTSFTransformerForecaster(seq_len, context_len, pred_len, *, num_epochs=16, batch_size=8, in_channels=1, individual=False, criterion=None, criterion_kwargs=None, optimizer=None, optimizer_kwargs=None, lr=0.001, custom_dataset_train=None, custom_dataset_pred=None, position_encoding=True, temporal_encoding=True, temporal_encoding_type='linear', d_model=512, n_heads=8, d_ff=2048, e_layers=3, d_layers=2, factor=5, dropout=0.1, activation='relu', freq='h')[source]

LTSF-Transformer Forecaster.

Implementation of the Long-Term Short-Term Feature (LTSF) transformer forecaster, aka LTSF-Transformer, by Zeng et al [1]_.

Core logic is directly copied from the cure-lab LTSF-Linear implementation [2]_, which is unfortunately not available as a package.

Parameters:

seq_lenint: Length of the input sequence. Preferred to be twice the pred_len.
context_lenint, optional (default=2): Length of the label sequence. Preferred to be same as the pred_len.
pred_lenint: Length of the prediction sequence.
num_epochsint, optional (default=16): Number of epochs for training.
batch_sizeint, optional (default=8): Size of the batch.
in_channelsint, optional (default=1): Number of input channels.
individualbool, optional (default=False): Whether to use individual models for each series.
criterionstr or callable, optional: Loss function to use.
criterion_kwargsdict, optional: Additional keyword arguments for the loss function.
optimizerstr or callable, optional: Optimizer to use.
optimizer_kwargsdict, optional: Additional keyword arguments for the optimizer.
lrfloat, optional (default=0.001): Learning rate.
custom_dataset_traintorch.utils.data.Dataset, optional: Custom dataset for training.
custom_dataset_predtorch.utils.data.Dataset, optional: Custom dataset for prediction.
position_encodingbool, optional (default=True): Whether to use positional encoding. Positional encoding helps the model understand the order of elements in the input sequence by adding unique positional information to each element.
temporal_encodingbool, optional (default=True): Whether to use temporal encoding. Works only with DatetimeIndex and PeriodIndex, disabled otherwise.
temporal_encoding_typestr, optional (default=”linear”): Type of temporal encoding to use, relevant only if temporal_encoding is True. - “linear”: Uses linear layer to encode temporal data. - “embed”: Uses embeddings layer with learnable weights. - “fixed-embed”: Uses embeddings layer with fixed sine-cosine values as weights.
d_modelint, optional (default=512): Dimension of the model.
n_headsint, optional (default=8): Number of attention heads.
d_ffint, optional (default=2048): Dimension of the feedforward network model.
e_layersint, optional (default=3): Number of encoder layers.
d_layersint, optional (default=2): Number of decoder layers.
factorint, optional (default=5): Factor for the attention mechanism.
dropoutfloat, optional (default=0.1): Dropout rate.
activationstr, optional (default=”relu”): Activation function to use. Defaults to relu and otherwise gelu.
freqstr, optional (default=”h”): Frequency of the input data, relevant only if temporal_encoding is True.

Attributes:

cutoff: Cut-off = “present time” state of forecaster.
fh: Forecasting horizon that was passed.
is_fitted: Whether fit has been called.
state: State of the estimator.

Examples

>>> from sktime.forecasting.ltsf import LTSFTransformerForecaster
>>> from sktime.datasets import load_airline
>>>
>>> y = load_airline()
>>>
>>> model = LTSFTransformerForecaster(10, 5, 5)
>>> model.fit(y, fh=[1, 2, 3, 4, 5])
LTSFTransformerForecaster(context_len=5, pred_len=5, seq_len=10)
>>> pred = model.predict()

Methods

`build_pytorch_pred_dataloader`(y, fh)	Build PyTorch DataLoader for prediction.
`build_pytorch_train_dataloader`(y)	Build PyTorch DataLoader for training.
`check_is_fitted`([method_name])	Check if the estimator has been fitted.
`clone`()	Obtain a clone of the object with same hyper-parameters and config.
`clone_tags`(estimator[, tag_names])	Clone tags from another object as dynamic override.
`create_test_instance`([parameter_set])	Construct an instance of the class, using first test parameter set.
`create_test_instances_and_names`([parameter_set])	Create list of all test instances and a list of names for them.
`fit`(y[, X, fh])	Fit forecaster to training data.
`fit_predict`(y[, X, fh, X_pred])	Fit and forecast time series at future horizon.
`get_class_tag`(tag_name[, tag_value_default])	Get class tag value from class, with tag level inheritance from parents.
`get_class_tags`()	Get class tags from class, with tag level inheritance from parent classes.
`get_config`()	Get config flags for self.
`get_fitted_params`([deep])	Get fitted parameters.
`get_param_defaults`()	Get object's parameter defaults.
`get_param_names`([sort])	Get object's parameter names.
`get_params`([deep])	Get a dict of parameters values for this object.
`get_pretrained_params`([deep])	Get pretrained parameters of this estimator.
`get_tag`(tag_name[, tag_value_default, ...])	Get tag value from instance, with tag level inheritance and overrides.
`get_tags`()	Get tags from instance, with tag level inheritance and overrides.
`get_test_params`([parameter_set])	Return testing parameter settings for the estimator.
`get_y_true`(y)	Get y_true values for validation.
`is_composite`()	Check if the object is composed of other BaseObjects.
`load_from_path`(serial)	Load object from file location.
`load_from_serial`(serial)	Load object from serialized memory container.
`predict`([fh, X])	Forecast time series at future horizon.
`predict_interval`([fh, X, coverage])	Compute/return prediction interval forecasts.
`predict_proba`([fh, X, marginal])	Compute/return fully probabilistic forecasts.
`predict_quantiles`([fh, X, alpha])	Compute/return quantile forecasts.
`predict_residuals`([y, X])	Return residuals of time series forecasts.
`predict_var`([fh, X, cov])	Compute/return variance forecasts.
`pretrain`(y[, X, fh])	Pre-train forecaster on panel (global) data.
`reset`()	Reset the object to a clean post-init state.
`save`([path, serialization_format])	Save serialized self to bytes-like object or to (.zip) file.
`score`(y[, X, fh])	Scores forecast against ground truth, using MAPE (non-symmetric).
`set_config`(**config_dict)	Set config flags to given values.
`set_params`(**params)	Set the parameters of this object.
`set_random_state`([random_state, deep, ...])	Set random_state pseudo-random seed parameters for self.
`set_tags`(**tag_dict)	Set instance level tag overrides to given values.
`update`(y[, X, update_params])	Update cutoff value and, optionally, fitted parameters.
`update_predict`(y[, cv, X, update_params, ...])	Make predictions and update model iteratively over the test set.
`update_predict_single`([y, fh, X, update_params])	Update model with new data and make forecasts.