Skip to content
ML Academy

Classical ML

Interview question

  • stafflevel
  • 25 minsuggested time
  • 2Data scientist, ML engineer

A bank's small-business credit team uses a logistic-regression scorecard. A new gradient-boosting model improves ROC-AUC from 0.74 to 0.79 on a holdout. Risk, compliance and the business all have opinions. How would you decide whether to replace the scorecard, and what would the decision process look like?

Tests: Model selection beyond accuracy: business value of a metric gain, explainability and fairness requirements, stability, monitoring, governance, and designing a decision process across stakeholders.

day 1day 11day 12day 13day 58

Sign in to practise: answers are saved before model answers are shown.