Expose one-hot encoding to H2OFrame operations

[exalate-issue-sync[bot]]

Requesting a method on a H2ODataframe with one categorical column that outputs a multi column dataframe (one column for each unique category) and a 0/1 value for that row.

[exalate-issue-sync[bot]]

Vlad Patryshev commented: It's in branch vlad_PUBDEV_3955, waiting for Michal's approval.

[exalate-issue-sync[bot]]

Michal Malohlava commented: It will be closed, when issue hit the master.

[exalate-issue-sync[bot]]

Erin LeDell commented: Let’s revisit this. Lots of interest on Stack Overflow, especially by people using interpretability methods.

[exalate-issue-sync[bot]]

Erin LeDell commented: [~accountid:557058:04659f86-fbfe-4d01-90c9-146c34df6ee6] just pointed out that R has a (private, non-exported) function which already does this FYI {{.getExpanded}}: [https://github.com/h2oai/h2o-3/blob/master/h2o-r/h2o-package/R/frame.R#L312|https://github.com/h2oai/h2o-3/blob/master/h2o-r/h2o-package/R/frame.R#L312|smart-link]

[exalate-issue-sync[bot]]

Michal Kurka commented: [~accountid:557058:afd6e9a4-1891-4845-98ea-b5d34a2bc42c] this function is GLM-specific - it might produce a different order of columns than how xgboost sees the frame - which might be important in some cases

[h2o-ops]

JIRA Issue Migration Info

Jira Issue: PUBDEV-3955 Assignee: Vlad Patryshev Reporter: Mark Chan State: Reopened Fix Version: N/A Attachments: N/A Development PRs: Available

Linked PRs from JIRA

https://github.com/h2oai/h2o-3/pull/898 https://github.com/h2oai/h2o-3/pull/957