Post written by Rachael Rho:

H2O Flow

I wanted to share an interesting, H2O Flow hands-on workshop I participated in with the H2O Open Tour conference last week. H2O Flow is a notebook-style open-source user interface for H2O, similar to iPython Notebook or RStudio. Many thanks to Amy for walking us through this intro!

If you haven’t already, you’ll need to install H2O and its necessary dependencies indicated here. Then in your H2O directory, you can execute the following in terminal or command prompt to access Flow at http://localhost:54321.

java -jar h2o.jar


You should see a page similar to the below! Again, H2O Flow looks to be a pretty interactive user interface where you can easily import and explore data, as well as build models on the H2O system.

Flow Initial Page Screenshot

During the session, we implemented a few machine learning models on an Airlines dataset. While I can’t share the dataset here, below is both an example load, logit model build and finally, results and out-of-the-box model visualization:

Flow Steps Screenshot

Flow Model Results Screenshot

Flow seems pretty easy to use – though obviously you can always still connect to H2O with Python or R. Overall, I was pretty impressed that H2O took the time to walk the whole conference through the installation process to model results and hope this was also interesting to you!