AWS Certification Exam AI Practitioner (Practice Questions) set7

問題 4 / 65

What important points should be considered when adjusting the threshold for imbalanced data?

The threshold should always be fixed at 0.5.
Lowering the threshold will always improve accuracy.
Threshold adjustment is unnecessary because it does not affect the model structure.
You can complete threshold selection by looking at the ROC curve alone.
Depending on the objective, set the threshold based on the PR curve and costs, taking into account the trade-off between precision and recall.

当サイトでは、ユーザー体験の向上を目的としてCookieを使用しています。サイトの利用を継続することで、Cookieの使用に同意したものとみなされます。