TimeSeriesDataFrame.train_test_split

TimeSeriesDataFrame.train_test_split(prediction_length: int, end_index: int | None = None, suffix: str | None = None) → Tuple[TimeSeriesDataFrame, TimeSeriesDataFrame][source]

Generate a train/test split from the given dataset. This method can be used to generate splits for multi-window backtesting.

Parameters:

  • prediction_length (int) – Number of time steps in a single evaluation window.

  • end_index (int, optional) – If given, all time series will be shortened up to end_idx before the train/test splitting. In other words, test data will include the slice [:end_index] of each time series, and train data will include the slice [:end_index - prediction_length].

  • suffix (str, optional) – Suffix appended to all entries in the item_id index level.

Returns:

  • train_data (TimeSeriesDataFrame) – Train portion of the data. Contains the slice [:-prediction_length] of each time series in test_data.

  • test_data (TimeSeriesDataFrame) – Test portion of the data. Contains the slice [:end_idx] of each time series in the original dataset.