pdstools.explanations.Plots

Classes

Plots

Module Contents

class Plots(explanations: pdstools.explanations.Explanations.Explanations)

Bases: pdstools.utils.namespaces.LazyNamespace

Parameters:

explanations (pdstools.explanations.Explanations.Explanations)

dependencies = ['numpy', 'plotly']
dependency_group = 'explanations'
X_AXIS_TITLE_DEFAULT = 'Contribution'
Y_AXIS_TITLE_DEFAULT = 'Predictor'
explanations
aggregate
contributions(top_n: int = defaults.top_n, top_k: int = defaults.top_k, descending: bool = defaults.descending, missing: bool = defaults.missing, remaining: bool = defaults.remaining, sort_by: str = defaults.sort_by.value, display_by: str = defaults.display_by.value)

Plots contributions for the overall model or a selected context.

Args:
top_n (int):

Number of top predictors to display.

top_k (int):

Number of top unique values for each categorical predictor to display.

descending (bool):

Whether to sort the predictors by most or least contribution.

missing (bool):

Whether to include missing values in the plot.

remaining (bool):

predictors/predictor values not included in the top_n/top_k will be grouped into a “remaining” category.

sort_by (str):

Type of contribution calculation to use for sorting/selecting top predictors. Default is contribution_abs to select most impactful predictors.

display_by (str):

Type of contribution calculation to display in plots. Default is contribution to show signed contribution values.

Returns:
tuple[go.Figure, list[go.Figure]]:

  • left: context header if context is selected, otherwise None

  • right: overall contributions plot and a list of predictor contribution plots.

Parameters:
plot_contributions_for_overall(top_n: int = defaults.top_n, top_k: int = defaults.top_k, descending: bool = defaults.descending, missing: bool = defaults.missing, remaining: bool = defaults.remaining, sort_by: str = defaults.sort_by.value, display_by: str = defaults.display_by.value) → tuple[plotly.graph_objects.Figure, list[plotly.graph_objects.Figure]]
Parameters:
Return type:

tuple[plotly.graph_objects.Figure, list[plotly.graph_objects.Figure]]

plot_contributions_by_context(context: dict[str, str], top_n: int = defaults.top_n, top_k: int = defaults.top_k, descending: bool = defaults.descending, missing: bool = defaults.missing, remaining: bool = defaults.remaining, sort_by: str = defaults.sort_by.value, display_by: str = defaults.display_by.value) → tuple[plotly.graph_objects.Figure, plotly.graph_objects.Figure, list[plotly.graph_objects.Figure]]
Parameters:
Return type:

tuple[plotly.graph_objects.Figure, plotly.graph_objects.Figure, list[plotly.graph_objects.Figure]]

static _build_hover_customdata(df: polars.DataFrame, x_col: str)

Build customdata array and hovertemplate for contribution plots.

Expects df to contain a frequency_pct column.

Returns (customdata, hovertemplate) with columns: predictor_name, predictor_type, contribution (x_col value), frequency_pct.

Parameters:

  • df (polars.DataFrame)

  • x_col (str)

_plot_overall_contributions(df: polars.DataFrame, x_col: str, y_col: str, x_title: str = X_AXIS_TITLE_DEFAULT, y_title: str = Y_AXIS_TITLE_DEFAULT, context: pdstools.explanations.ExplanationsUtils.ContextInfo | None = None) → plotly.graph_objects.Figure
Parameters:
Return type:

plotly.graph_objects.Figure

_plot_predictor_contributions(df: polars.DataFrame, x_col: str, y_col: str, x_title: str = X_AXIS_TITLE_DEFAULT, y_title: str = Y_AXIS_TITLE_DEFAULT) → list[plotly.graph_objects.Figure]
Parameters:

  • df (polars.DataFrame)

  • x_col (str)

  • y_col (str)

  • x_title (str)

  • y_title (str)

Return type:

list[plotly.graph_objects.Figure]

static _plot_context_table(context_info: pdstools.explanations.ExplanationsUtils.ContextInfo) → plotly.graph_objects.Figure
Parameters:

context_info (pdstools.explanations.ExplanationsUtils.ContextInfo)

Return type:

plotly.graph_objects.Figure