Checkpoint: Beyond the Basics

Question 1 of 5

• A balanced mix of all columns.
• The count columns, because there are more of them.
• Bands of annual spend, because it dominates the distances.
• Random groupings, because k-means can’t handle mixed units.

Question 2 of 5

• The LLM’s number is more accurate because the model is larger.
• There is no difference; both are model outputs.
• predict_proba can’t be used for ranking.
• The logistic regression’s probabilities were trained and tested against actual churn outcomes; the LLM’s number is generated text that hasn’t been.

Question 3 of 5

• Demographic parity (equal selection rates).
• Equal opportunity (equal true positive rates).
• Calibration, because recall is a calibration measure.
• None; fairness only concerns selection rates.

Question 4 of 5

• Do nothing until outcomes prove performance has dropped.
• Delete mobile_share from the model immediately.
• Investigate the shift, check flag rates per group, and plan a retrain and re-audit once enough new labelled data exists.
• Lower the threshold to compensate.

Question 5 of 5

• Compress the file so it can’t be modified.
• Rename it to .pkl so tools recognise it.
• Load it with the newest scikit-learn release available.
• Load only files from your own access-controlled model store, with the scikit-learn version recorded at training time.