topic Re: Values to look for in Score node in SAS Data Science

Values to look for in Score node

Babloo — Wed, 14 Jun 2017 12:06:34 GMT

I wanted to score a model using SAS Eminer. But I'm not certain about the values to understand whether the model has a good score.

Appreciate if you could shed some light on this.

Re: Values to look for in Score node

WendyCzika — Fri, 16 Jun 2017 15:23:37 GMT

Do you mean you want to assess a model where the target is known, or you want to score new data to create predictions where you don't have a target?

Re: Values to look for in Score node

Babloo — Fri, 16 Jun 2017 16:39:39 GMT

I want to score a new data where the target is unknow.

##- Please type your reply above this line. Simple formatting, no
attachments. -##

Re: Values to look for in Score node

WendyCzika — Fri, 16 Jun 2017 17:46:34 GMT

You can look at the results of the modeling node or use a Model Comparison node to see how well that model fits the training data (and validation/test data if you have those partitions). Then if you connect a score data set to the Score node along with the model you want to score with, then you will get new column(s) containing the predictions using that model. But I'm not sure if I'm answering your question?

Re: Values to look for in Score node

Babloo — Fri, 16 Jun 2017 17:54:39 GMT

What value we need to look in the results? Higher the value better the
model?

##- Please type your reply above this line. Simple formatting, no
attachments. -##

Re: Values to look for in Score node

WendyCzika — Fri, 16 Jun 2017 19:21:35 GMT

For most of the Fit Statistics (misclassification rate, anything with "error" in it), smaller is better. In ROC and Lift plots, higher is better.