Using the grid search meta-estimator with the "mean_square_error", the "mean_absolute_error", the "median_absolute_error" or the "log_loss" as scoring parameters leads to the negation of those metrics. This is confusing especially for new users.
I suggest that we prefix those strings by "neg-" or "negative_". This would make clear from the start that the score is obtained from the negation of the loss / error.
Using the grid search meta-estimator with the
"mean_square_error", the"mean_absolute_error", the"median_absolute_error"or the"log_loss"as scoring parameters leads to the negation of those metrics. This is confusing especially for new users.I suggest that we prefix those strings by
"neg-"or"negative_". This would make clear from the start that the score is obtained from the negation of the loss / error.