Exercise: Equal-Interval Chart
Learn to visualize model performance with an equal-interval chart.
We'll cover the following...
Using intervals of predicted probability
In this exercise, you'll make a plot similar to the "Default rate according to model prediction decile" plot from the last lesson. However, instead of splitting the test set into equal-population deciles of predicted probability, you'll use equal intervals of predicted probability. Specifying the intervals could be helpful if a business partner wants to think about potential model-based strategies using certain score ranges. You can use pandas cut
to create equal-interval binnings, or custom binnings using an array of bin edges, similar to how you used qcut
to create quantile labels.
Perform the following steps to complete the exercise:
-
Create the series of equal-interval labels, for 5 bins, using the following code:
equal_intervals, equal_interval_bin_edges = \ pd.cut(x=test_set_pred_proba,\ bins=5,\ retbins=True)
Notice that this is similar to the call to
qcut
, except here withcut
we can say how many equal-interval bins we want by supplying an integer to thebins
...