Working with the model explainability interface

The model explainability interface is a simple function that incorporates various graphs and information about the model and its workings. There are two main functions for model explainability in H2O:

• The h2o.explain() function, which is used to explain the model’s behavior on the entire test dataset. This is also called global explanation.
• The h2o.explain_row() function, which is used to explain the model’s behavior on an individual row in the test dataset. This is also called local explanation.

Both these functions work on either a single H2O model object, a list of H2O model objects, or the H2O AutoML object. These functions generate a list of results that consists of various graphical plots such as a variable importance graph, partial dependency graph, and a leaderboard if used on multiple models.

For graphs and other visual results, the explain object relies on visualization engines to render the...