PMML code- categorical values

[ace2131]

HI

I have a PMML model which is created using IBM SPSS. It has optype=“categorical” while KNIME doesn’t uses this datatype as SPSS defines.

Columns/Features used in the model contain numbers but they are typified as Nominal and Ordinal in IBM SPSS and PMML shows them as Categorical.

I did look at some of the forum threads which mention to convert to string and then use domain calculator but when I do so, I get an error → column “ABC” in the table of prediction is expected to be numeric. If I don’t convert to string and then use domain calculator, I get an error- The column “ABC” is supposed to be nominal.

Can anyone suggest a solution?

Thanks

[julian.bunzel]

Hi @ace2131,

is it possible for you to share a screenshot of your workflow?
Which node do you use to do the predictions?

Best,
Julian

[ace2131]

attached

[mlauber71]

You could try the approaches in this example, compiling the pmml or converting it into SQL

Model Deployment SQL Recoding – iris

Also you might take a sample dataset like the iris or census income data and build a model with that so you could share the pmml here so we could further explore.