The model building and model scoring contexts
In Section 2, Building State-of-the-Art Models on Large Data Volumes Using H2O, we spent a great amount of focus on building world-class models at scale with H2O. Building highly accurate and trusted models against massive datasets can potentially generate millions of dollars for a business, save lives, and define new product areas, but only when the models are deployed to production systems where predictions are made and acted upon.
This last step of deploying and predicting (or scoring) on a production system can often be time-consuming, problematic, and risky for reasons discussed shortly. H2O makes this transition from a built (trained) model to a deployed model easy. It also provides a wide range of flexibility in regard to where scoring is done (device, web application, database, microservice endpoint, or Kafka queue) and to the velocity of data (real-time, batch, and streaming). And, whatever the production context, the H2O deployed...