topic How to create ROC curves and Lift curves in SAS Programming

How to create ROC curves and Lift curves

WhiteDiamond — Tue, 25 Oct 2016 06:56:33 GMT

Dear Community,

I'm using the Base SAS 9.2 in order to replicate the findings of a study made using R tool and rminer package. The file that I'm currently using, Bank Marketing Dataset, is a dataset that has been created from a Portuguese Bank to predict if the client will subscribe a term deposit or not.

I've been searching for a solution on the Internet, and all the analysis that I've found on this dataset are made using R. Up to this point, I already analyzed the main variables and attributes using SAS; however I'm not able to calculate the ROC curves and the Lift curves for the models: Logistic Regression, Decision Tree and Support Vector Machine in order to determine which is the most appropriate model to make the prediction.

It would be great if someone in these community could give me any suggestion. Thanks for your time.

Re: How to create ROC curves and Lift curves

Ksharp — Tue, 25 Oct 2016 10:09:01 GMT

I guess you should use Decision Tree and Random Forest .

Re: How to create ROC curves and Lift curves

Rick_SAS — Tue, 25 Oct 2016 13:05:23 GMT

Although you say "Base SAS," I assume you mean SAS/STAT procedures.

Try limiting your search to the proceedings of SAS conferences to get targeted SAS-only solutions. For example, do an internet search for

roc lift site:lexjansen.com

There are many results, some going back 20 years.

You should also consider updating your version of SAS. SAS 9.2 is from March 2008, so it's almost nine years old. There have been seven releases of SAS since then, with an eighth release coming out shortly. Newer releases include procedures like HPSPLIT which enable you to form tree-based regression models with SAS/STAT. PROC LOGISTIC also contains the ROC statement, which provides a convenient way to work with ROC curves.