MAINT Use class_of_interest in DecisionBoundaryDisplay

[ArturoAmorQ]

Since scikit-learn v1.4 DecisionBoundaryDisplay accepts class_of_interest for multiclass visualization. This feature is promised in the current version of the MOOC.

Notice that, as it requires updating the minimal version, it may change the experience of current enrolled participants.

[glemaitre]

I think that we can simplify the code by using some groupby that will avoid using the loc here:

from matplotlib import cm

_, axs = plt.subplots(ncols=3, nrows=1, sharey=True, figsize=(12, 5))
plt.suptitle("Predicted probabilities for decision tree model", y=1.05)
plt.subplots_adjust(bottom=0.45)

for idx, (class_of_interest, data_class) in enumerate(
    data_test.groupby(by=target_test)
):
    axs[idx].set_title(f"Class {class_of_interest}")
    disp = DecisionBoundaryDisplay.from_estimator(
        tree,
        data_test,
        response_method="predict_proba",
        class_of_interest=class_of_interest,
        ax=axs[idx],
        vmin=0,
        vmax=1,
    )
    data_class.plot.scatter(
        x="Culmen Length (mm)",
        y="Culmen Depth (mm)",
        ax=axs[idx],
        marker="o",
        s=100,
        c="w",
        edgecolor="k",
    )

ax = plt.axes([0.15, 0.14, 0.7, 0.05])
plt.colorbar(
    cm.ScalarMappable(cmap="viridis"), cax=ax, orientation="horizontal"
)
_ = plt.title("Probability")

[glemaitre]

I feel that we have less matplotlib boilerplate code in this case since we rely on pandas directly.