We are building a Hadoop-based predictive analytics application. A number of Random Forests are built and a PMML file of each is generated. At the very end, we want to aggregate all forests in one final PMML file so that the emitted prediction is the average of predictions by the forests. We call this a Meta-Forest (a forest of Random Forests). Is this supported by/doable in PMML? I tried nested segmentation where each inner segment contains a whole segmentation block of an individual RF, but it seems not to work.