Only Plot Nominal Variable separately in Sensitivity Analysis plot (#…

…2381)

Summary:
Pull Request resolved: #2381

In Sensitivity Analysis plot, we want to plot (unordered) categorical separately and shows that it "affects" metrics while still showing "increases"/"decreases" metrics for ordinal variables.

Reviewed By: dme65

Differential Revision: D56325368

fbshipit-source-id: 9e9e271d61f07f0e4c5fc952eced8d2c530aad4d

ax/plot/feature_importances.py

@@ -149,7 +149,7 @@ def plot_feature_importance_by_feature_plotly(
         categorical_features = [
             name
             for name, par in model.model_space.parameters.items()
-            if isinstance(par, ChoiceParameter)
+            if isinstance(par, ChoiceParameter) and not par.is_ordered
         ]
 
     for i, metric_name in enumerate(sorted(sensitivity_values.keys())):