BookmarkSubscribeRSS Feed
🔒 This topic is solved and locked. Need further help from the community? Please sign in and ask a new question.
Joel_Wesley
Fluorite | Level 6

Hello,

The ROC curve (and various other output) aren't always showing up when I open the results to the model comparison node.  Instead I only get Fit Statistics and Output.  Occasionally it shows up, but I can't seem to figure out why it does in some diagrams/flows but not in others.  Has anyone else had the same problem?  I'm working in EM 12.1.

Thanks!

Jo

1 ACCEPTED SOLUTION

Accepted Solutions
Joel_Wesley
Fluorite | Level 6

Just to update this thread - I received help a while ago from SAS tech support.  Seems as though I shouldn't have been including segment variables - i.e., I should have been "rejecting" them at the data loading stage.  I was keeping them, but calling them "segment."  SAS tech also helped me use the "advanced" editor on data loading stages, which I wasn't trying before.  Anyway, problem solved - ROC curves are showing up consistently now!

View solution in original post

9 REPLIES 9
WendyCzika
SAS Employee

When it's not showing up, do you have a binary target and the property ROC Chart set to Yes?

Joel_Wesley
Fluorite | Level 6

Yes, those are always my settings.  I'm beginning to think that EM simply won't calculate ROC when the sample becomes too large.  I suppose that makes sense, as it involves running through the data repeatedly with many different cutoffs.  Still, I have obtained ROC curves in base SAS with far larger samples.  This won't seem to do it with ~3000.

WendyCzika
SAS Employee

Hmm, there are no sample size limitations that I'm aware of.  This sounds like one for SAS Tech Support.  Can you please  create a model package with the flow where you are seeing this issue (see 46764 - Creating a model package (SPK file) in SAS® Enterprise Miner(tm)) and send it to Tech Support for assistance: http://support.sas.com/ctx/supportform/createForm.

Joel_Wesley
Fluorite | Level 6

Sounds good.  Thanks for your help!

YujunChiang
Calcite | Level 5

perhaps you only need to set the role of target variable from "interval" to "binary". I found that when  target variable has value 0 and 1, the default role is not "binary", and I can't see roc before I set the role to "binary".

Joel_Wesley
Fluorite | Level 6

Thanks, it is set it to binary, but I still find it not producing ROC curves.

Joel_Wesley
Fluorite | Level 6

Just to update this thread - I received help a while ago from SAS tech support.  Seems as though I shouldn't have been including segment variables - i.e., I should have been "rejecting" them at the data loading stage.  I was keeping them, but calling them "segment."  SAS tech also helped me use the "advanced" editor on data loading stages, which I wasn't trying before.  Anyway, problem solved - ROC curves are showing up consistently now!

M_Maldonado
Barite | Level 11

Very useful update Joel!

Thank you!

bcastillo10
Calcite | Level 5

Thanks for the details and the update. 

Ready to join fellow brilliant minds for the SAS Hackathon?

Build your skills. Make connections. Enjoy creative freedom. Maybe change the world. Registration is now open through August 30th. Visit the SAS Hackathon homepage.

Register today!
How to choose a machine learning algorithm

Use this tutorial as a handy guide to weigh the pros and cons of these commonly used machine learning algorithms.

Find more tutorials on the SAS Users YouTube channel.

Discussion stats
  • 9 replies
  • 6998 views
  • 6 likes
  • 5 in conversation