Rapid Recall Key 21–40

1. Batch: process complete/recomputed historical data.
2. Speed: process very recent/real-time data for low latency.
3. Serving: expose/merge batch and speed results for consumers.
4. Trade-off: completeness vs latency, with synchronization/complexity/cost.
5. Supervised: learn generalized rules from examples with known/labeled targets/outcomes.
6. Unsupervised: discover hidden patterns/groupings without target labels.
7. Reinforcement: learn actions through feedback/reward toward a goal.
8. Fast discriminator: does a known/labeled target exist?
9. Profiling: examine structure, content, patterns, distributions, and characteristics to understand a source.
10. Reduction: reduce variables/data complexity while retaining analytically important information.
11. Association vs clustering: find elements/events that occur together vs group similar observations/cases.
12. Self-organizing map: organize/map complex similarity into a view that reveals groups/relationships.
13. Sentiment: infer attitude/opinion/emotional polarity or related meaning from text in context.
14. Keyword weakness: negation, context, sarcasm, and word combinations change meaning.
15. Mining vs predictive: discover patterns/relationships vs estimate likelihood of a specified future outcome.
16. Predictive vs operational: estimate an outcome vs apply analytics/models to current operational streams, possibly triggering alerts/actions.
17. Source criteria: business relevance, origin, format, meaning, relationships, update frequency, grain, consistency, reliability, foundational value, privacy, bias, feasibility/value.
18. Granularity: different detail levels cannot be safely compared/joined until grain is understood and aligned.
19. Timing: a source can be too stale/volatile for the decision; update cycle must fit the use.
20. Filtering bias: source-selection rules can systematically exclude populations/events, making evidence unrepresentative.

← Prompts 21–40 · Key 41–60 →