topic Re: Using logistic regression in SAS/EG like a task in a workflow in SAS Enterprise Guide

Using logistic regression in SAS/EG like a task in a workflow

juanvg1972 — Mon, 30 Nov 2015 21:37:39 GMT

Hi,

I have a process made in SAS/BASE coding that I want to do now in SAS/EG using workflow.

I have to do a logistic regression. In my below code you can see that I run a logistic regression (training dataset), save the model and after I apply my model with other dataset (validation)

proc logistic data=test outmodel=modelo plots=ROC;
class cod_tarifa productos;
model hc_consumo=cod_tarifa productos;
quit;

/* Apply model*/

proc logistic inmodel=modelo;
score data=validar out=validar1;
quit;

I am trying to do it with logistic regression task in a workflow. I can select target variable (hc_consumo) and features (cod_tarifa productos), but I don't know how to save the model, and how to apply after hist model, is it possible in a SAS/EG guided task.

Any advice will be greatly appreciated

Juan

Re: Using logistic regression in SAS/EG like a task in a workflow

ChrisHemedinger — Tue, 01 Dec 2015 12:52:21 GMT

Hi Juan,

I don't think that the Logistic Regression task supports the SCORE statement.

There is this task: Tasks->Data Mining->Model Scoring. This allows you to run a stored model against another data set, but the model must be registered in a SAS Metadata repository. (The task is available only if you are connected to a SAS Metadata environment.) That is, you would need to use something like SAS Model Manager or SAS Enterprise Miner to have registered your model from a previous step, then you can select the model to use with your validation data.

Otherwise, you would need to do what you're doing: build the model in the Logistic Regression task, but then use a SAS program node to score it.

Others who have expertise in data mining might have a better answer.

Chris

Re: Using logistic regression in SAS/EG like a task in a workflow

juanvg1972 — Tue, 01 Dec 2015 13:30:52 GMT

Thanks for your help, Chris

I will try model scoring tasks. Previosuly I would like to save the model, is it possible from the Logistic Reg task of SAS/EG.

I think it is not posiible, but I am not sure.

Thanks,