F1-score
Why the harmonic mean rather than the average, what F1 cannot see, and when a single balanced number is the wrong thing to want.
Why the harmonic mean rather than the average, what F1 cannot see, and when a single balanced number is the wrong thing to want.
The same model at the same operating point can score 0.97 or 0.03 on precision. What changed was the prevalence, not the model.