TFT (Temporal Fusion Transformer) is a deep learning model that combines LSTM and attention mechanisms to capture complex temporal patterns and forecast multiple time series.
Training parameters
Number of validation iterations before early stopping, -1 for unlimited.
The size of the batches to be used for training and prediction.
Using a fixed random seed allows for reproducible result.
Model parameters
Whether the points with a hidden size above a specific threshold should be skipped during the hyperparameter search.