plot_residuals#

plot_residuals(forecast_df: DataFrame, ts: TSDataset, feature: str | Literal['timestamp'] = 'timestamp', transforms: Sequence[Transform] = (), segments: List[str] | None = None, columns_num: int = 2, figsize: Tuple[int, int] = (10, 5))[source]#

Plot residuals for predictions from backtest against some feature.

Parameters:

  • forecast_df (DataFrame) – forecasted dataframe with timeseries data

  • ts (TSDataset) – dataframe of timeseries that was used for backtest

  • feature (str | Literal['timestamp']) – feature name to draw against residuals, if “timestamp” plot residuals against the timestamp

  • transforms (Sequence[Transform]) – sequence of transforms to get feature column

  • segments (List[str] | None) – segments to use

  • columns_num (int) – number of columns in subplots

  • figsize (Tuple[int, int]) – size of the figure per subplot with one segment in inches

Raises:

ValueError: – if feature isn’t present in the dataset after applying transformations

Notes

Parameter transforms is necessary because some pipelines doesn’t save features in their forecasts, e.g. etna.ensembles pipelines.