Skip to content

CS Professional · Artificial Intelligence, Data Analytics and Cyber Security - Laws and Practice · Artificial Intelligence - Introduction and Basics

A bank's fraud-detection AI performs well in testing but, after six months, its accuracy drops sharply because customer spending behaviour and fraud tactics have changed. Which is the most accurate diagnosis and response?

This is model drift. Real-world behaviour and fraud patterns changed after deployment, so the model's learned relationships no longer hold. The remedy is continuous monitoring and retraining or recalibration with fresh data, not faster hardware or discarding all historical data.

  1. AModel drift; monitor performance continuously and retrain or recalibrate the model with fresh dataCorrect
  2. BOverfitting; the fix is to stop using any historical data
  3. CData poisoning by the bank's own staff; the fix is to dismiss the data team
  4. DHardware obsolescence; the fix is to buy faster processors only

Explanation

When real-world data patterns change after deployment, a model's accuracy degrades; this is model (concept) drift. Governance requires ongoing monitoring and retraining. Faster hardware does not fix changed data patterns, and nothing indicates malicious poisoning.

Did you get it right without looking?

One question tells you little. A timed set on Artificial Intelligence - Introduction and Basics shows your real accuracy, how long you take and where you lose marks.

Practise Artificial Intelligence - Introduction and BasicsFree diagnostic mock

More Artificial Intelligence - Introduction and Basics questions