Temporal Fusion Transformer (TFT)#

class oats.models.predictive.tft.TFTModel(window: int = 10, n_steps: int = 1, use_gpu: bool = 1, val_split: float = 0.2, **kwargs)[source]#

Bases: DartsModel

TFT Model (Temporal Fusion Transformer)

Using TFT as a predictor. Anomalies scores are deviations from predictions.

Reference: https://unit8co.github.io/darts/generated_api/darts.models.forecasting.tft_model.html

initialization also accepts any parameters used by: https://unit8co.github.io/darts/generated_api/darts.models.forecasting.tft_model.html

Parameters:

  • window (int, optional) – rolling window size to feed into the predictor. Defaults to 10.

  • n_steps (int, optional) – number of steps to predict forward. Defaults to 1.

  • use_gpu (bool, optional) – whether to use GPU. Defaults to False.

  • val_split (float, optional) – proportion of data points reserved for validation. Defaults to 0.2.

fit(train_data: ndarray[Any, dtype[Any]], epochs: int = 15, **kwargs)#
get_scores(test_data: ndarray[Any, dtype[float32]], normalize=False, **kwargs) → Tuple[ndarray[Any, dtype[float32]]]#