topic Re: SAS Enterprise Miner in SAS Data Science

SAS Enterprise Miner - using display values vs numeric variables for model training

mmaccora — Tue, 29 Aug 2017 13:16:26 GMT

Hi,

My question is the following:

Consider a data set where we have 2 variables called "Fruit" and another one called "CFruit" with corresponding codes.

Ex:
Variable "Fruit": lemon, orange, melon, banana, lemon etc.
Code Variable for Fruit "CFruit": 1, 7, 3, 5, 1, etc.

So that, 1 = lemon, 7 = orange, 3 = melon, 5 = banana, etc.

In SAS Enterprise Miner, is it possible to use the "CFruit" variable as input to train a model (for instance a decision tree), but display the corresponding labeled variable "Fruit" ?

In fact, it is easier for me to interpret the model with labels instead of codes, but for implementing the process in an automated chain it is more convenient and secured to use codes.

Thank you very much for your help.

Best Regards,
Marco

Re: SAS Enterprise Miner

Reeza — Sat, 26 Aug 2017 16:34:34 GMT

Create a format and apply the format to the variable in the SAS data set.

It should operate as you desire. It's probably easier to do that in your prep step, which I find easier outside of EM.

Re: SAS Enterprise Miner

mmaccora — Tue, 29 Aug 2017 05:24:53 GMT

Thank you very mich for your answer it works!

Is it possible now to bypass the hard coded format ?

What I would like is to based my format on another column "one-to-one":

Dataset:
1, lemon
7, orange
3, melon
5, banana
1, lemon

A code like:

proc format;
value $LabelFruit.
CFruit = Fruit;
run;

Do you know a way of doing such kind of formatting ?

Re: SAS Enterprise Miner

Reeza — Tue, 29 Aug 2017 06:38:53 GMT

When in doubt check the documentation, especially the examples.

http://support.sas.com/documentation/cdl/en/proc/70377/HTML/default/viewer.htm#n1e19y6lrektafn1kj6nbvhus59w.htm