topic Re: AUC and ALIFT in SAS Data Science

AUC and ALIFT

Amitkr — Mon, 14 Dec 2020 02:20:02 GMT

I am a beginner and trying a comparison of models. I have a few basic questions:

1. Is AUC same as ROC index, if not then how can I get AUC?

2. What is ALIFT in model comparison result?

(**NOTE** I found both (AUC and ALIFT) terms used in a research paper for model comparison, now I am building models to compare my results with these results.)

3. Do we need to encode non-numerical data of the input file before building a data mining model in SAS?

Thank You

Re: AUC and ALIFT

Ksharp — Mon, 14 Dec 2020 12:50:39 GMT

Attachement is lift polt.

And the following code is getting ROC curve.

data have;
set sashelp.class(rename=(weight=score));
good_bad=rand('bern',0.8);
keep good_bad score ;
run;

proc logistic data=have  ;
model good_bad=score/ nofit ;
roc 'Roc Curve1' pred=score;
run;

Re: AUC and ALIFT

Amitkr — Mon, 14 Dec 2020 13:36:58 GMT

Thanks for responding 🙂

I am also getting the lift, ROC plot in my results. What I am not clear about is whether AUC and ROC index are the same things and what is ALIFT (not lift).

and one more doubt I have is whether we need to encode(e.g., one-hot encoding, label encoding etc.) all categorical variables before importing data in SAS or it is handled by SAS automatically.

Hope I am clear now.

Re: AUC and ALIFT

Ksharp — Tue, 15 Dec 2020 12:15:36 GMT

Yes. I think "AUC and ROC index are the same things " both stand for the area under curve .
ALIFT I have no clue . maybe @Rick_SAS know it.
And categorical variables are handled by SAS automatically(CLASS statement). no need encode before modeling.

Re: AUC and ALIFT

Amitkr — Tue, 15 Dec 2020 12:19:05 GMT

Thank You 🙂