Module 4 Assessment: Probabilistic Classifiers and Calibration Check

Estimated active time: 30-45 minutes

Question 1

What does a calibration table show that accuracy does not?

Pass answer: explains that it tests whether a stated probability matches the observed rate.

Question 2

Report your buckets. Is the model over- or under-confident?

Pass answer: gives predicted versus actual per bucket with counts, and reads the direction correctly.

Question 3

Why is a bucket with five rows weak evidence?

Pass answer: recognises that a rate over few observations moves in large steps.

Question 4

When does calibration matter more than accuracy?

Pass answer: identifies threshold-driven decisions where the probability itself is used.