Skip to content
ML Academy

Forecasting

Interview question

  • seniorlevel
  • 15 minsuggested time
  • 2Data scientist, ML engineer

Your model's nominal 80% prediction intervals for weekly route demand cover the actual value only 62% of the time across 500 backtest forecast origins. Planners use the upper bound to set capacity. Is the 62% noise? How do you fix it, and how do you check the fix?

Tests: Evaluating interval calibration, understanding why intervals under-cover, and conformal or empirical-residual repairs with time-aware validation.

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