Module 07 Knowledge Check
5 questions. Pass mark 4 out of 5. Answer every question before checking the answer key below, then retry after reading the feedback.
1. On imbalanced data, accuracy primarily reports…
- A. model quality
- B. the class balance
- C. the threshold
- D. the training time
2. The two error types in a confusion matrix are…
- A. always equally costly
- B. usually very differently costly, which is a business fact
- C. interchangeable
- D. only relevant for balanced data
3. A threshold of 0.5 is…
- A. always optimal
- B. correct only when both errors cost the same
- C. required by most frameworks
- D. the maximum-recall point
4. Errors are most informative when sorted by…
- A. file name
- B. confidence, highest first
- C. date
- D. class
5. A slice with 40 images should be reported as…
- A. a reliable accuracy figure
- B. unmeasured - too few rows to be stable
- C. the model's worst case
- D. an outlier to exclude
---
Answer Key and Explanations
Check these only after attempting every question.
1. B - the class balance
A constant predictor can beat a model that finds half the positives.
2. B - usually very differently costly, which is a business fact
The cost ratio determines the operating point and is not a modelling decision.
3. B - correct only when both errors cost the same
Where a miss costs 45x a false alarm, the cheapest threshold is well below 0.5 - and nothing about the model changed.
4. B - confidence, highest first
A confident mistake means a learned rule is wrong; a borderline one means the case was hard.
5. B - unmeasured - too few rows to be stable
Report row counts with every slice metric; 62% over 40 images and over 4,000 are different claims.
Practical Check
Apply this module to your own work: complete the module activity for *Classification Evaluation and Error Analysis*, then write one sentence naming what your result shows and one naming what it does not.
Strong Answer Pattern
A strong answer names the task, the evidence used, the check performed, and the remaining limitation. It avoids "proved", "guaranteed", or "always" unless the evidence genuinely supports it.