Skip to content
ML Academy

Classical ML

Interview question

  • midlevel
  • 10 minsuggested time
  • 1Data scientist

You fit a logistic regression for flight cancellations on bootstrap resamples of the same data. Two features, distance_km and scheduled_minutes, correlate at 0.98. The coefficient of distance_km is positive in some resamples and negative in others, yet the model's AUC barely moves. Is the model broken? What do you do?

Tests: Recognising multicollinearity as an inference problem rather than a prediction problem, quantifying it (VIF), and choosing a remedy that fits the purpose of the model.

day 11day 17

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