Module 08 Self-Test

A: Predicting majority class always = high accuracy, zero usefulness. Use F1, PR-AUC, or recall depending on cost of false positives vs false negatives.

A: Creates synthetic minority samples by interpolating between existing ones. ONLY apply to training data — never test/validation (data leakage).

A: Subtracting an incrementing row number from consecutive dates produces the same group value. Non-consecutive dates produce different values, creating separate groups you can count.

A: Pivot: aggregate + reshape to see a metric across two dimensions. Melt: convert column headers into row values (wide → long format).

A: Consistency, Availability, Partition tolerance.

A: No — "fail to reject H₀" ≠ "H₀ is true." It means insufficient evidence at this sample size.