4

How is the score of GridsearchCV calculated? Is the score a percentage? Does this mean higher is a better?

Ethan
  • 1,625
  • 8
  • 23
  • 39
ml_learner
  • 347
  • 1
  • 4
  • 11

1 Answers1

4

The score is based on the scorer defined in the scoring argument. Meaning, the scorer can be any of the default metrics, such as precision, accuracy or F1-score (e.g., this); or a custom scorer. For a scorer (by convention), higher value is better. The value is not necessarily a percentage, but is often normalized between 0 and 1.

Shayan Shafiq
  • 1,012
  • 4
  • 11
  • 24
Felix Z.
  • 172
  • 5
  • I create a Gradient Boost Regressor with a GridSearchcv but dont define the score. Whta does the score mean by default? – ml_learner Feb 11 '20 at 13:43
  • 2
    If None, the estimator’s score method is used. This means in your case the Gradient Boost Regressor default scorer is used (that is coefficient R^2) – Felix Z. Feb 11 '20 at 13:48
  • and r² is not a percentage? – ml_learner Feb 11 '20 at 14:00
  • 2
    By convention, we use word `score` when higher is better. We use word `loss` when less is better. When you don't have any particular in mind, use `metric`. – Piotr Rarus Feb 11 '20 at 14:53
  • 1
    @ml_learner r2 is not percentage (and it's not r-squared). It's 'kinda' normalized, but you can use it to compare models just fine. 1.0 - is perfect model 0.0 - this is constant answer model <0.0 - model goes wrong – Piotr Rarus Feb 11 '20 at 14:54
  • @PiotrRarus-ReinstateMonica I didn't know about the score / loss convention, but seems logic. Changed the answer accordingly. – Felix Z. Feb 11 '20 at 16:39