Plotter.attributes_obs_vs_pred

Plotter.attributes_obs_vs_pred(block: str = None, attributes: List[str] = None, transformed: bool = False, per_column: bool = False, downsamp: int = 1, n_cols: int = 1, datamodule: DataModule | None = None) → Figure | None[source]

Plots the observed (true) vs predicted values for all the selected attributes (= DataObject’s) in the given data block of the datamodule (InputML or OutputML). For numerical variables, plots a 2D scatter plot. For categorical variables, plots a confusion matrix.

Parameters:

  • block (str, optional, default=None) – Name of the data block to be plotted. The block must be available in the dataset or in the datamodule. If None, the block is automatically selected based on the attributes.

  • attributes (List[str], optional, default=None) – List of attributes (= names of DataObject’s) to be plotted. If None, all attributes are plotted. If the block argument is not None, the attributes need to be provided.

  • transformed (bool, optional, default=False) – If True, the data is plotted transformed, while False returns the plots it in its original domain

  • per_column (bool, optional, default=False) – If True, multi-dimensional attributes are plotted as separate traces and subplots, otherwise all data is flattened to a single trace.

  • downsamp (int, optional, default=1) – Down-sampling factor for the data.

  • n_cols (int, optional, default=1) – Number of columns in the plot.

  • datamodule (DataModule, optional, default=None) – A DataModule object used for evaluation. If None, the datamodule provided when creating the plotter is used.

Returns:

Optional[plotly.graph_objects.Figure] – Plotly figure object, if self.output is None, otherwise None.

Notes

The plot is generated using the validation set of the datamodule. Legends of the plot might not be displayed correctly in certain environments, see plotly/plotly.py#4489 for more information.