Logistic regression produces a conditional probability p rather than a class label directly. For evaluating performance
Posted: Tue May 17, 2022 2:18 pm
Logistic regression produces a conditional probability p
rather than a class label directly. For evaluating
performances of such probabilistic models, what strategy would be
the best? In other words, do you think that accuracy is the best
evaluation metric? If not, what metric or strategy should we
consider?
rather than a class label directly. For evaluating
performances of such probabilistic models, what strategy would be
the best? In other words, do you think that accuracy is the best
evaluation metric? If not, what metric or strategy should we
consider?