CERTIFIED-MACHINE-LEARNING-PROFESSIONAL Question #40: Real Exam Question with Answer & Explanation

Question

Which of the following is a simple, low-cost method of monitoring numeric feature drift?

Options

• AJensen-Shannon test
• BSummary statistics trends
• CChi-squared test
• DNone of these can be used to monitor feature drift
• EKolmogorov-Smirnov (KS) test

Explanation

Summary statistics trends (B) are correct because tracking simple metrics like mean, median, standard deviation, min/max, and percentiles over time requires no statistical test infrastructure - just logging and comparing numbers. It's immediately interpretable, computationally cheap, and works well as a first-pass alert for drift.

Why the distractors are wrong:

• A (Jensen-Shannon) measures divergence between two probability distributions - valid for drift detection, but computationally heavier and requires full distribution comparison, not "simple and low-cost."
• C (Chi-squared) tests for differences in categorical distributions, not numeric/continuous features - wrong data type.
• E (Kolmogorov-Smirnov) is a legitimate and powerful numeric drift test, but it's a formal statistical test requiring reference distributions and threshold tuning - more complex and costly than summary statistics.
• D is a trap answer; multiple methods can detect drift, so "none" is always suspicious here.

Memory tip: Think "cheap = descriptive." Summary statistics are descriptive stats you'd compute in any EDA - no hypothesis testing machinery needed. If the question says simple and low-cost, it's pointing away from formal statistical tests toward basic monitoring dashboards.

No community discussion yet for this question.