Skip to main content


Performs Temporal Fusion Transformer forecasting

Source Code: [link]

openbb.forecast.tft(data: Union[pd.Series, pd.DataFrame], target_column: str = "close", n_predict: int = 5, past_covariates: str = None, train_split: float = 0.85, forecast_horizon: int = 5, input_chunk_length: int = 14, output_chunk_length: int = 5, hidden_size: int = 16, lstm_layers: int = 1, num_attention_heads: int = 4, full_attention: bool = False, dropout: float = 0.1, hidden_continuous_size: int = 8, n_epochs: int = 200, batch_size: int = 32, model_save_name: str = "tft_model", force_reset: bool = True, save_checkpoints: bool = True)


data (Union[pd.Series, pd.DataFrame])Input DataNoneTrue
target_columnOptional[str]:Target column to forecast. Defaults to "close".closeTrue
n_predict (int, optional)Days to predict. Defaults to 5.NoneTrue
train_split (float, optional)Train/val split. Defaults to 0.85.NoneTrue
past_covariates (str, optional)Multiple secondary columns to factor in when forecasting. Defaults to None.NoneTrue
forecast_horizon (int, optional)Forecast horizon when performing historical forecasting. Defaults to 5.NoneTrue
input_chunk_length (int, optional)Number of past time steps that are fed to the forecasting module at prediction time.
Defaults to 14.
output_chunk_length (int, optional)The length of the forecast of the model. Defaults to 5.NoneTrue
hidden_size (int, optional)Hidden state size of the TFT. Defaults to 16.NoneTrue
lstm_layers (int, optional)Number of layers for the Long Short Term Memory Encoder and Decoder. Defaults to 16.NoneTrue
num_attention_headers (int, optional)Number of attention heads. Defaults to 4.NoneTrue
full_attention (bool, optional)Whether to apply a multi-head attention query. Defaults to False>NoneTrue
dropout (float, optional)Fraction of neurons affected by dropout. Defaults to 0.1.NoneTrue
hidden_continuous_size (int, optional)Default hidden size for processing continuous variables. Defaults to 8.NoneTrue
n_epochs (int, optional)Number of epochs to run during training. Defaults to 200.NoneTrue
batch_size (int, optional)Number of samples to pass through network during a single epoch. Defaults to 32.NoneTrue
model_save_name (str, optional)The name for the model. Defaults to tft_modelNoneTrue
force_reset (bool, optional)If set to True, any previously-existing model with the same name will be reset
(all checkpoints will be discarded). Defaults to True.
save_checkpoints (bool, optional)Whether or not to automatically save the untrained model and checkpoints from training.
Defaults to True.


Adjusted Data series,
List of historical fcast values,
List of predicted fcast values,
Optional[float] - precision,
Fit Prob. TFT model object.