Why do we use negative scoring and then negate it again? Could we just use positive scoring?
I think the creators of Scikit-learn use negative scoring as a trick to have an easier classification since the worst score will be the smaller when you use negative scoring.
1 Like
Indeed, it linked to this idea.
In search CV, you are interested in either maximizing a score (greater is better) or minimize the error (lower is better). Scikit-learn made the choice to take only the score convention: we always maximize metrics. The question is thus: How do you transform an error to a metric → You negate it.
1 Like