Creating a model using Regression node

Posts: 0

Creating a model using Regression node

hello all,

i am doing a project of HEALTH CARE INTELLIGENCE using DATA Mining.

I Want to predict the Outcome of the Patient depending on the previous data i have. And this Target variable is "Patient Disposition" which is number from 1-20 each number indicating each outcome like 1--Home,2--Transfer Hospital etc etc...

When i provide the SCORE data with other fields as input but the Patient disposition field goes into UNARY managed and it gets rejected. And its not allowing me to assign it as target.

I make use of REGRESSION node in this process..

How to convert this field to Target?

Please Help ME its urgent.....
Trusted Advisor
Posts: 2,114

Re: Creating a model using Regression node

I'm not an SAS Enterprise Miner user, so my response is more toward general SAS procedures.

What you are describing is a logistic regression with a nominal outcome. SAS LOGISTIC will not do that analysis, it can only do binary or ordinal logistic regressions.

Even if Enterprise Miner can do nominal logistic regression, it may not be able to create a stable model using the SEMMA approach. To get stable estimates requires about 15 observations per potential predictor for the rarest outcome. With 20 outcomes, you will run up against a sample size limitation on the SEMMA limit of 32,000 obserations.

You are probably going to have to collapse your outcomes and either get to an ordinal scale or a series of binaries.
Posts: 0

Re: Creating a model using Regression node

No its going into Linear Regression Method.. as the Target variable in Input node is INTERVAL. But in Input node were we give the input dataset as SCORE to Regression node there the target is turning into UNARY.
Ask a Question
Discussion stats
  • 2 replies
  • 2 in conversation