BookmarkSubscribeRSS Feed
rogelio_mancisidor
Calcite | Level 5

Hi,

 

I have import some bins into my IGN using the option Import Data Set in the property panel. In addition, I have set to YES the option Use Frozen Groupings, since I have done some changes afterwards. For some reason, IGN cannot find any split for some variables. This must be an error because if I run another IGN from scratch, those variables are grouped.

 

So, is it possible to kind of 'reset ' a variable and so the IGN will group it again from 'scracth'? since I have chosen to import the grouping data, will I need to delete the variable that I would like to 'reset' from the imoprted table?

 

Thanks

1 REPLY 1
JasonXin
SAS Employee
rogelio_mancisidor, First, thanks for using SAS. My name is Jason Xin, advanced analytics solution architect working at SAS Institute. I just saw your post. When the Use Frozen Groupings property is set to Yes and you run the Interactive Binning node, automatic grouping is not performed on variables where grouping definitions have already been created. You therefore do not have 'collision' or 'involuntary override', within one single IG node. Automatic grouping /training based on data on variables other than those performed by the frozen groupings will play out alongside. It is not uncommon to observe that imposed frozen groupings do not yield as expected. Sometimes they are cutting out of the range of (new) data being tested. Sometimes they give you surprising distributions. These kinds of 'symptoms' are usual diagnosis modelers run to see if a rebuild or recalibration of the model is called for. Visually inspecting grouped variables is one matter. Checking their accuracy or predictive potential is another. If you are interested to carry along the groupings generated by the frozen set with the machine trained groupings, you may consider using two IG nodes (they don't even need to be immediately after the other). One processes the frozen and the other gives you the automatic. Check imported and exported data sets to make sure they are not overriding each other. They should not. Hope this helps. Happy holiday. Best Regards Jason Xin

SAS Innovate 2025: Save the Date

 SAS Innovate 2025 is scheduled for May 6-9 in Orlando, FL. Sign up to be first to learn about the agenda and registration!

Save the date!

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
  • 1 reply
  • 1147 views
  • 0 likes
  • 2 in conversation