Getting Probabilities in PMML output for Random Forest

Hi All,
Hi @AlexanderFillbrunn,

I have done the Random forest & i have exported that output in PMML.

One thing i noticed that, in the PMML output file i an getting the decision boundaries (either 0 or 1 as probability). Please help me here as i want to know the actual probability (probability of being 1) how can i do that? When i looked at Multiple Models Method it has options like sum, average, max vote etc not sure how to get the probability

Please refer below screen-shot:
Capture1

Hi @ChetanP,
unfortunately the Table to PMML Ensemble node cannot produce a PMML that can predict probabilities. Do you need the PMML format, or can you also just save the model of the Random Forest Learner using a Model Writer node?

2 Likes

Hi @AlexanderFillbrunn,

Thanks for your reply.

I am looking for PMML format so that i can put that directly into production. Also, if i save the model using a Model Writer then how can i put that in production? Please help me on this.

Is there any other ways wherein i can save the model in PMML format that has probabilities in it and put that in production?

Please suggest

1 Like

You can store every model either with the model writer or there is a separate writer. As long as you stay on the KNIME platform you could just re use it.

One more possiblity is the H2O environment. The MOJO format can be used on a local machine but also on a big data server with the help of sparkling water.

Maybe if you tell us more about the production environment we will be able to help you.

1 Like

This topic was automatically closed 182 days after the last reply. New replies are no longer allowed.