Treatment of boundary for values for cut-off

pvareschi — Sun, 03 May 2020 06:25:10 GMT

When applying a cut-off to classify a case (by default set at 0.5), how does Enterprise Miner treat a case with a posterior probability equal to exactly 0.5? For example, "Lesson 4: Decision Trees Using SAS Enterprise Miner" -> "Working with Decision Trees" (under "Applied Analytics Using SAS Enterprise Miner") does not cover this eventuality. However, "Lesson 7: Model Assessment Using SAS Enterprise Miner", "Calculating Expected Profit" seems to hint to the fact that scenario would be classified as a primary event (i.e. 1).

Does the same logic applies when creating the data for a ROC chart? ("Lesson 7: Model Assessment Using SAS Enterprise Miner", "ROC charts")

Re: Treatment of boundary for values for cut-off

Cynthia_sas — Mon, 04 May 2020 13:24:46 GMT

Hi:

Feedback from the instructors:

"SAS EM nodes by default classifies the case as an event when the posterior probability > or eq to 0.5. If the posterior probability is < 0.5 then the case is classified as non-event."

Hope this helps,

Cynthia

topic Treatment of boundary for values for cut-off in SAS Academy for Data Science

Treatment of boundary for values for cut-off

Re: Treatment of boundary for values for cut-off