I just took to learn Meteor and notice that ether I cannot find any, or there is no integration packages for Machine learning with Meteor applications. Only thing I could find was prediction.io package that is updated last time august 2014. Could someone help to right direction on this.
Also I like to hear people’s opinion whether there should be package for this or is every one fine for just passing transaction information trough REST API and using something like prediction.io, WEKA or one of the ML- as a service options ?
I am looking mostly supervised learning use case. For example user X with known to read and like specified content (features) so user Y with similar taste (features) is presented same content while user Z with different preferences is shown different content.
Why I would like to use something like prediction.io event server instead of just implementing this is mongo models is performance and prediction accuracy.
I can build something like this it in native javascript (probably in meteor as well soon as I get my head around it) but most optimised ML libraries exist in Python and Java. These allow near real time speed, and I like idea of near realtime predictions on meteor application content.
Now without any pre-made integration how I would build this is first deploy prediction.io as web service then use HTTP Rest to sent data and receive prediction as JSON this JSON content will then fire a template update to generate updated content view. https://docs.prediction.io/deploy/
Well, to report various UI interaction events to cloud there is astronomer.io project.
And if you have some heavy suggestion engine planned as me, u will be better with neo4j as database for that social relation network part.
I am atm modeling my neo4j relational model to search above all livestreaming services and suggestions/discovery mode based on channel likes/follows etc.
Thanks shock that medium article is brilliant resource! Initially I am planning only to capture few events but I probably end up with many more including users social relationships, and the there is nothing better than graph database to capture and understand those event relationships.