I built out a series of linear regressions fed by a single “table to H20” node + H20 Local Content node" yesterday, so I can demonstrate to clients a progression of steps.
I thought that the “H20 Local Content” node was directing the KNIME processes to save directly to my local folder. Now I have a complete system that does not work at all as it is looking for data on the cloud. I had planned on using for presentations many of these “cascading” style analyses using many different H20 analyses nodes, but can’t have errors like this embarrassing me when nothing works. What is going wrong?
Could you provide us with more information about your settings and your system and what this workflow does look like. The screenshot is very small and you cannot see much details about waht is going on.
In general H2O.ai would create its own environment (based on Java) and would only exist as long as your WF is in and executed state and you cannot just save and restore it.
After separating the clusters I use the H20 table to feed several different linear regressions on each clustered group to show clients how i adjust the independent variables progressively
then i start the series of regressions you see in the original screen shot.
I want to do this for several different regression nodes.
My concern is if H20 does not store and run locally on my computer, and if i don’t have internet for a presentation, my models may simply give me the error messages and make me appear foolish to the client.
you don’t need internet connection if you are using the “H2O Local Context” node. It basically creates a “cloud” locally on your machine. You just need to re-execute the node when closing and re-opening the workflow because the H2O frames (the yellow connections) are gone when closing the workflow.
