Import PMML model to Model Manager

[valengvz]

Hi there, I'm working with migration from R to SAS, the company currently has a model in R which has written with the h2o package. I know for experience that Model Manager accepts PMML models so I already did the conversion from MOJO(type of file that h2o's model export) to PMML. 

When I imported to SAS I had the following message error:

This is how the PMML file looks like: 

<?xml version="1.0" encoding="UTF-8" standalone="yes"?>
<PMML xmlns="http://www.dmg.org/PMML-4_4" xmlns:data="http://jpmml.org/jpmml-model/InlineTable" version="4.4">

I don't know how can I fix this error because when I tried to load the PMML model to python the model works fine and I even made predictions.

Does someone know how can I deal with this error or if there is another alternative to work with h2o models from R in SAS?

[GregDolland-EDC]

We faced a similar issue and found out that SAS Viya only supports PMML v4.2, not PMML v4.4. The following community link discussed how to create a properly formatted PMML 4.2 for R, though you might find that PMML 4.2 might not support the model you're trying to deploy.

https://communities.sas.com/t5/SAS-Nordic-Users-Group/Juletip-14-Importing-R-model-to-SAS-Viya-with-...

If this doesn't work, your options would be to

• recreate your model using SAS code and deploy that
• publish your R package to a supported container service

more details on the PMML implementation are in the SAS Model Manager documentation

https://documentation.sas.com/doc/en/mdlmgrug/15.3/p0t1l5scv87yoqn1m13rzx6yy45g.htm

Best of luck! 

[valengvz]

I already did that, that works with other models but not with gradient boosting or random forest. I does SAS support ensemble models via pmml?