PublicationsFrom: Foster Provost fprovost@stern.nyu.eduDate: Tue, 13 Feb 2001 17:00:36 -0500 Subject: Provost on dealing with the case when one class is rare I think the lesson in the last two paragraphs in Gary Weiss email (GPS: see previous item) is very important. Researchers have done experiments modifying the class distribution, and have not recalculated the decision threshold (calling into question the results). I imagine that practitioners also might find this useful. |
Copyright © 2001 KDnuggets. Subscribe to KDnuggets News!