Create a segmented model training interface in R

[exalate-issue-sync]

Proposal to change the way we generate "H2OSegmentModels" in R (currently it duplicates all modeling functions e.g. h2o.gbm() -> h2o.bulk_gbm() and adds two segment related parameters, and returns "H2OSegmentModel"):

It seems like the closest thing to what we’re doing with the “bulk train” is grid search. With grid search, we have a singular generic function h2o.grid() where you can apply that to any of the algos, and you just pass along args as needed. So in that design paradigm, in R, we would do the bulk/segment training as something like:

{code:r}models <- h2o.segment_train(algorithm = "gbm", segment_models_id = "foo", segment_columns,
x, y, training_frame, ...){code}

Not sure exactly what the function name should be, but the important part is that the API closes matches h2o.grid(). Just the two additional segment params, and then we can just pass all the other arg names down the …

Note: Make sure to remove all the old code, including the python scripts that auto-generated it.

[h2o-ops]

JIRA Issue Migration Info

Jira Issue: PUBDEV-7422
Assignee: Erin LeDell
Reporter: Erin LeDell
State: Resolved
Fix Version: 3.30.0.1
Attachments: N/A
Development PRs: Available

Linked PRs from JIRA

#4480
#4477