Skip to content
ML Academy

Forecasting

Interview question

  • midlevel
  • 8 minsuggested time
  • 2Data scientist, Data analyst

A colleague evaluated a daily passenger forecasting model with shuffled 5-fold cross-validation and reports 3% MAPE. What is wrong, and how would you evaluate it?

Tests: Time-aware validation: rolling-origin backtesting, baselines, per-horizon error, and the information available at each forecast origin.

day 15day 23day 25

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