H2O Model to MOJO Using Group Loop Categories
This is just on my local machine 5.4.4. I found a way to Group Loop 74 separate categories through the H2O Random Forest Regression valuation system. I can see how the valuations are indeed customized for each category and how the Variable Importance measures of the IVs adjust for each category.
I thought then that the Learner and Predictor had then created and stored the custom valuation model for each of the 74 categories.
But when I try to pull forecast data into the H2O MOJO Predictor (Regression) via an Excel File, it returns the exact same predictions for every category. The previously learned and customized modeling does not seem to apply at the MOJO level.
I thought perhaps because I used a Group Loop to create the initial H2O models that Looping each category through the MOJO would work. But having a second loop blows up the system.
The WF and all data is attached. Thoughts: 1) Did I really create 74 unique models–the 74 different 90%+ R2 scores seem to suggest so, and the 74 different Feature Importance scores seem to suggest so; and 2) if so, how can I run each of 74 models through the MOJO one-at-a-time?
Thank you