Validating Your H2O Model Using Deepchecks

Deepchecks now supports an out-of-the-box integration with H2O AutoML models.

Introduction

AutoML tools have been increasingly adopted by Data Science teams in the past few years. Although many aspects of the modeling process are still firmly in the domain of the individual Data Scientist, for some use-cases AutoML tools can bring significant value in the modeling stage in a fraction of the time it’ll take to manually create a good model.

That’s why Deepchecks now natively supports running its checks and suites on models created with the H2O AutoML library – one of the most popular AutoML tools out there. Once you created your H2O model, you can simply wrap it with this Deepchecks wrapper:

from deepchecks.tabular.integrations.h2o import H2OWrapper

h2o_deepchecks_model = H2OWrapper(h2o_model)

And go on to run any Deepchecks check or suite using the regular Deepchecks API. For example:

from deepchecks.tabular.suites import model_evaluation

model_evaluation().run(train_dataset, test_dataset, h2o_deepchecks_model)

The added bonus? Our h2o_deepchecks_model object is now a fully sklearn-compatible model, including built-in feature importance! You can go on and use this Deepchecks wrapper to run any sklearn metric, model selection tool or other utils on this model.

Next Steps

For additional information you can follow the full tutorial, showing how to train a regular sklearn model and an H2O model for the same task, and comparatively evaluate their performance using deepchecks. Go try it out yourself!

Don’t forget to ⭐ our Github repo, it’s really a big deal for open-source-led companies like Deepchecks.