2

I am very curious to know how H2O can help in Fraud Ring Detection. I am referring to Venkatesh Reamanthan's video here where he talks about fraud ring detection and wanted to know if there are any samples published by H2O.

We all know that, if we know a fraud scenario, we can easily construct a Neo4J Cipher query to detect fraud rings. I am not able to understand how a graph db can be fed to a ML model and when a new transaction happens how ML model can predict if the parties in the transaction are part of a fraud ring or not?

My understanding is the below after going through the video : If we have a graphdb, if I somehow export this to an edgelist and feed to Node2Vec to produce a word2vec model, and then feed the output / dataset extracted out of this word2vec model to H2O driverless AI for further training. Is my understanding correct? Are there any samples to do it?

At this point, I am clue less as to how to convert Node2Vec's output in to a format that h2o driverless AI can understand and how can it actually predict a fraud ring.

Any help, pointers, ideas appreciated.

SRK
  • 169
  • 1
  • 2
  • 13
  • the video you are referencing is using Driverless AI one of the products of H2O.ai the company. The tag you specified for this post is `h2o` which refers to h2o-3 H2O.ai's open source platform. Can you clarify whether you are feeding your data to a H2O-3 model or a Driverless AI model, by updating the title and the question content. Note both of these products expect that a dataset, not model, will be fed to them. – Lauren Aug 13 '18 at 17:15
  • @Lauren made the edits as suggested by you. – SRK Aug 14 '18 at 06:18
  • driverless-ai expects tabular data, you can see the list of expected data formats here: http://docs.h2o.ai/driverless-ai/latest-stable/docs/userguide/release_notes.html – Lauren Aug 15 '18 at 18:57
  • @SRK Have you found any relevant lectures/examples with graph db and H2o since you asked? :) – mbh86 Dec 10 '18 at 11:56

0 Answers0