Skip to content

Classification Drill — 20 Fast Decisions

Classify before looking at the key.

  1. A model estimates next-month default probability.
  2. The system recommends the credit-limit action expected to reduce loss.
  3. Billions of rows overwhelm current storage/processing.
  4. Sensor events must be acted on within seconds.
  5. Two files use different identifiers and structures and are hard to combine.
  6. A provider's data cannot be trusted.
  7. Labeled fraud/not-fraud examples train a classifier.
  8. No labels; the team wants natural customer groups.
  9. An agent improves choices using reward feedback.
  10. Find products frequently purchased together.
  11. Group similar customers without predefined segments.
  12. A raw-data repository has no source/provenance/inventory and users cannot find data.
  13. A source is well documented but 40% of a critical key is missing.
  14. A source is complete but nobody knows origin or lineage.
  15. A model is tuned every time the “test” score changes.
  16. Training accuracy is high but unseen-data performance is poor.
  17. Rare transactions are the possible fraud cases being sought.
  18. A plant can act only once per day, but a vendor proposes millisecond analytics.
  19. Two anonymous sources together identify small groups.
  20. Petabytes of mixed raw logs need cheap flexible landing, not immediate complex SQL.

Key

  1. Predictive analytics
  2. Prescriptive analytics
  3. Volume
  4. Velocity; likely speed/operational-analytics requirement
  5. Viscosity (Variety supports if formats differ)
  6. Veracity
  7. Supervised learning
  8. Unsupervised learning / clustering
  9. Reinforcement learning
  10. Association / data mining
  11. Clustering / unsupervised mining
  12. Data-swamp risk / Metadata failure
  13. Data Quality problem
  14. Metadata problem
  15. Test-set independence failure; use validation for tuning
  16. Over-fitting/generalization failure
  17. Do not auto-delete; investigate as possible target signal
  18. Latency/cost mismatch; business does not justify speed architecture
  19. Recombination privacy risk
  20. Distributed file-based environment

← Reconstructions · Repair Protocol →