LTSFNLinearForecaster

LTSF-NLinear Forecaster.

Quickstart

from sktime.forecasting.ltsf import LTSFNLinearForecaster

estimator = LTSFNLinearForecaster(seq_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)

Parameters(11)

seq_lenint

length of input sequence

pred_lenint

length of prediction (forecast horizon)

num_epochsint, default=16

number of epochs to train

batch_sizeint, default=8

number of training examples per batch

in_channelsint, default=1

number of input channels passed to network

individualbool, default=False

boolean flag that controls whether the network treats each channel individually” “or applies a single linear layer across all channels. If individual=True, the” “a separate linear layer is created for each input channel. If” “individual=False, a single shared linear layer is used for all channels.”

criteriontorch.nn Loss Function, default=torch.nn.MSELoss

loss function to be used for training

criterion_kwargsdict, default=None

keyword arguments to pass to criterion

optimizertorch.optim.Optimizer, default=torch.optim.Adam

optimizer to be used for training

optimizer_kwargsdict, default=None

keyword arguments to pass to optimizer

lrfloat, default=0.003

learning rate to train model with

Examples

>>> from sktime.forecasting.ltsf import LTSFNLinearForecaster
>>> from sktime.datasets import load_airline
>>> model = LTSFNLinearForecaster (10, 3)
>>> y = load_airline ()
>>> model. fit (y, fh = [1, 2, 3 ]) LTSFNLinearForecaster(pred_len=3, seq_len=10)
>>> y_pred = model. predict ()
>>> y_pred 1961-01 455.628082 1961-02 433.349640 1961-03 437.045502 Freq: M, Name: Number of airline passengers, dtype: float32

References

1. [1 ] Zeng A, Chen M, Zhang L, Xu Q. 2023. Are transformers effective for time series forecasting? Proceedings of the AAAI conference on artificial intelligence 2023 (Vol. 37, No. 9, pp. 11121-11128)... [R658c94d8ec96-2] https://github.com/cure-lab/LTSF-Linear