02 — Exam Map & High-Yield Targets
Chapter 14 is a 2% Fundamentals domain. Study it as a decision-and-process chapter, not as a vendor/algorithm trivia chapter.
24 targets
- Targets 01–08: foundations, V's, architecture, ML setup
- Targets 09–16: mining, sources, Metadata/DQ, alignment, model-vs-data quality
- Targets 17–24: model evidence, visualization/deploy, tools, governance/readiness
Highest-yield elimination rule
When two answers sound plausible, ask: 1. What business decision is being supported? 2. What lifecycle stage is the stem actually describing? 3. Is the problem source trust/context, DQ/alignment, model evidence, architecture, or operationalization? 4. Which changed fact would make the tempting distractor correct?
Ten changed-fact flips to rehearse
- churn probability → recommended retention action = predictive → prescriptive
- huge data → changes every second = volume → velocity becomes leading V
- managed lake loses inventory/lineage = lake → swamp risk
- labels disappear and natural groups are sought = supervised → unsupervised
- high training accuracy collapses on unseen data = over-fitting/generalization failure
- business can act daily, not per second = low-latency need may disappear
- daily/monthly sources become same grain = timing alignment issue removed
- anonymous sources identify a small group when combined = recombination privacy risk
- prototype begins triggering production actions = deployment/monitoring controls become primary
- reliable source systematically excludes population = bias/representativeness risk
Source boundary: pp. 471–501.