Enterprise Miner is not recognizing variables denoted as target

[miwilliams]

Hi all,

I feel like I am missing some sort of huge "aha" on this, but I simply don't know enough to help myself.  I am trying to run various models in SAS EM on a data set of 5 different variables (age ranges, sex, number of recipients, geographic region, and diagnosis).  This is the first time I have imported the data, which is an Excel spreadsheet.  All of the setup seemed to go through fine; it appears all lines (just under 1700) were imported, and there are no missing data. 

However, from the time I tried running decision tree, interactive decision tree, regression, and even an auto neural, the run fails, indicating I need at least one target variable.  I have tried designating each variable as such, but it is still not recognizing as intended, and I am at a loss.

I hope the screen shots below will help assist in seeing what I have done.  I appreciate any help in the mean time.  Thanks in advance!

[BrettWujek · Accepted solution]

Hello @miwilliams - what you are missing here is that for predictive modeling you need to specify one of your variables to be the "target" that you are trying to predict. This is completely separate from saying whether or not you want to "Use" a variable in a certain modeling node.

You specify a variable to be a target by setting the "Role" of a variable to target, which you can do on the node where the data source is defined at the beginning of the diagram. Select that node and select the "..." next to "Variables" and then for whichever variable you want to be the target, change the Role to Target.

This should solve your issue.

[miwilliams]

Thank you so much.  I didn't include screen shots of where I thought I had done this earlier in the diagram, but I apparently (wherever I did designate the target) did it in the wrong place.  Designating within the node, which was one of the screen shots I included, was a last ditch effort.