Hi,
for both questions I answered that the balanced_accuracy was better as that’s what i got.
See the following pic
However, the expected answer was exactly the opposite. Repeating the evaluation several times one sees that this is a product of fluctuations, or simply looking to the std would already hint that the gain is not significant.
Anyway it could be good to advise, in the questionnaire to force a random_state so that the results are reproducible or rephrase these questions in terms of significance?
Thanks,
Pedro