About the presence of missing values after predicting with a PMML Ensemble model

Hi guys,

I'm using the Tree Ensemble Learner node to model a biological endpoint, and then I export it to a PMML file using the following nodes: Tree Ensemble Model Extract, Table to PMML Ensemble, PMML writer.

In another workflow I use the PMML Reader and the PMML Ensemble Predictor nodes to predict the modeled endpoint for new molecules. Unfortunately after prediction it seems that 7% of molecules have a missing predicted value. Do you know what may be causing this? I have no clue.

Thanks in advance for your help.


Do you get the same predictions from the Tree Ensemble predictor node? 

Or maybe they one class isn't favoured and the selected option for a tie was missing?

Hello Swebb,

Thank you for your propt reply. You are completely right, I didn't realize I was having the "missing" option selected for a tie.