Building models with SAS Enterprise Miner, SAS Factory Miner, SAS Visual Data Mining and Machine Learning or just with programming

Specifying Levels, Roles, and Drop Codes in Enterprise Miner

Accepted Solution Solved
Reply
Frequent Contributor
Posts: 115
Accepted Solution

Specifying Levels, Roles, and Drop Codes in Enterprise Miner

PI am using Enterprise Miner (EM) Decision Trees to investigate a wide variety of models. To simplify things I have two models using the same dataset. I would like to use the same data and same data partition for both model runs. The only difference between the two models are a different Target variable for each.

It was highly recommended to set the Drop specifier to Y rather than set the Level specifier to Rejected. But the problem is that when I do this the variable literally gets rejected from my EM run and I cannot choose to change the Target variable to another one when I do my second model.

I hope that makes sense. Any recommendations?

Also, there are some variables that I want to still be included within my EM dataset for posterity, but there is no way I will use it within the model itself. So basically my variables are either Level = Target, Input, or Key. Can anyone suggest the best way to code these up so that there are no future problems down the line? I suppose I can set them to INPUT then set Drop = Y, and call it a day.

Thank you.


Accepted Solutions
Solution
‎12-08-2014 02:42 PM
SAS Super FREQ
Posts: 272

Re: Specifying Levels, Roles, and Drop Codes in Enterprise Miner

If I'm understanding correctly, I don't think you want to use Drop=Y in either of these situations.  That is removing the variable from the data set for all subsequent nodes.  For the two target situation, as long as you are not using Target1 as an input when modeling Target2 and vice-versa, you can set them both as Target variables, then in the modeling nodes, set Use=Yes to the one you want to use as the target and Use=No to the other.  If you do want to use them as inputs when modeling the other one, you can use a Metadata node to change the two targets between target/input and input/target, if that makes sense. 

For the other situation where you want to keep variables but not include them in the model, you can either set them as Role=Rejected to begin with, or as Role=Input with Use=No in whatever nodes you want to exclude them from.

View solution in original post


All Replies
Solution
‎12-08-2014 02:42 PM
SAS Super FREQ
Posts: 272

Re: Specifying Levels, Roles, and Drop Codes in Enterprise Miner

If I'm understanding correctly, I don't think you want to use Drop=Y in either of these situations.  That is removing the variable from the data set for all subsequent nodes.  For the two target situation, as long as you are not using Target1 as an input when modeling Target2 and vice-versa, you can set them both as Target variables, then in the modeling nodes, set Use=Yes to the one you want to use as the target and Use=No to the other.  If you do want to use them as inputs when modeling the other one, you can use a Metadata node to change the two targets between target/input and input/target, if that makes sense. 

For the other situation where you want to keep variables but not include them in the model, you can either set them as Role=Rejected to begin with, or as Role=Input with Use=No in whatever nodes you want to exclude them from.

☑ This topic is SOLVED.

Need further help from the community? Please ask a new question.

Discussion stats
  • 1 reply
  • 1371 views
  • 1 like
  • 2 in conversation