H2O allows “users to combine the fast, scalable machine learning algorithms of H2O with the capabilites of Spark” – sounds all very intersting. Having not had time to play with H2O it would appear that Flow is key to an almost like Clojure REPL (“Read-Eval-Print Loop”) environment. This in itself is interesting for the obvious reason of wanting a RAD environment to improve the models.
Fraud Detection is as usual one of the use cases offered by H2O. I suspect however that financial companies maybe interested in H2O given the issues in the compliance world, especially after the LIBRO incidents the sector has encountered over the last time period.
Customer Intelligence from my perspective would probably be interesting to Proof of Concept (PoC) in the client trading space – think RFQ’s, think trading patterns, think pushing dedicated research to clients etc
Anyone built any interesting PoC’s in H2O?