Decision 5 · Week 11

"What's unusual?"

Fraud, equipment failure, security breaches, medical events — the things that matter most are often the rarest. Anomaly detection finds the needles, not the haystack.

Anomaly Detection Market Basket Unsupervised

Why this decision matters

Most business value hides in the tails. A fraud rate of 0.5% — easy to ignore with accuracy as your metric — can be a multi-million-dollar opportunity. The same techniques that find fraud find equipment about to fail, customers about to leave, content going viral. We pair anomaly detection with market basket analysis — finding which items co-occur — because both are about spotting patterns in transactional data.

By the end of this topic you'll be able to

Distinguish point anomalies, contextual anomalies, and collective anomalies; choose a detection approach (statistical, distance-based, density-based, or model-based) for a given business problem; tune the trade-off between false alarms and missed events; mine transaction baskets to find item associations.

Materials

Key concepts to know
  • Three types of anomalies — point (one weird value), contextual (weird given context, e.g., AC running at 3am), collective (a sequence that's weird as a group).
  • Statistical methods — z-score, IQR, modified z. Quick, interpretable, assume distributions.
  • Distance / density methods — K-nearest-neighbors, isolation forest, local outlier factor. Don't assume a distribution.
  • Streaming detection — when you can't wait for the whole dataset; alerts in near-real-time.
  • Market basket analysis — support, confidence, lift. "People who bought X also bought Y."
  • Cost asymmetry — a missed fraud is worth $10,000; a false alarm is worth $5. Optimize accordingly.
Slides & class notes
Hands-on worksheet
Python notebooks
Datasets
Practice with games · Anomaly & market-basket games

Six anomaly-flavored games and one cross-topic escape room. Start with the MBA Student Guide for orientation, then try the detective games.

Sample presentations (executive vs. technical)

Stay Ahead of the Curve

Subscribe to our bi-weekly newsletter for the latest insights on AI, data, and business strategy.