topic Re: Problem with Scoring Regression Model at SAS Enterprise Miner in SAS Data Science

Problem with Scoring Regression Model at SAS Enterprise Miner

bryanssilveira — Fri, 10 Aug 2018 12:31:42 GMT

Hi there.

I am facing a problem to make predictions when I use Score in a Regression Model (this occur only if I use Regression).

The output of the Scoring is all equal, but the input's data are different. When I used other models, like Neural Network, this not occur.

So, how can I solve this problem?

Best Regards,

Bryan Silveira

Re: Problem with Scoring Regression Model at SAS Enterprise Miner

WendyCzika — Fri, 10 Aug 2018 14:44:07 GMT

If you look at the results of your Regression, do you see any effects being included in the model? I'm guessing it is resulting in an intercept-only model so that is why you are getting all the same predictions.

Re: Problem with Scoring Regression Model at SAS Enterprise Miner

bryanssilveira — Fri, 10 Aug 2018 16:13:31 GMT

The value of the intercept is different to the result of the predictions. I will put in the attachments what have been occuring.

Re: Problem with Scoring Regression Model at SAS Enterprise Miner

WendyCzika — Fri, 10 Aug 2018 16:52:54 GMT

I think it's because the values for Suco_Fr are all missing in your score data set, so it can't calculate a true prediction based on the regression model, so just uses the mean of the target from the training data for the prediction.

Re: Problem with Scoring Regression Model at SAS Enterprise Miner

Reeza — Fri, 10 Aug 2018 16:53:34 GMT

I think it's because of the missing values - specifically the ones that are all missing (Suco_Fr) have the same estimates.

PS. Please include your pictures directly into the posts (as below) rather than as an attachment, it's much easier to reference and respond and doesn't clutter up machines that way as well.

Re: Problem with Scoring Regression Model at SAS Enterprise Miner

bryanssilveira — Fri, 10 Aug 2018 18:15:15 GMT

The problem was really this.

Thanks!!