结合使用Apache Mahout和Ruby on Rails

Question

I have a ruby on rails application. I have the idea of implementing recommendations in the application. I came to know about Apache Mahout through stackoverflow. Now, If I have to use Mahout, what are the stuff that I have to do. Since it is a Java library, I am not clearly sure how to use it in an Ruby on Rails application. I have a couple of certifications in Java, so coding in Java is not an issue. I also have the idea of hosting the application in Heroku, would that be an issue?

Thanks

Answer 1

The recommender portion actually began life as a separate project which had nothing to do with Hadoop. It is still alive an well in Mahout. Look at everything under org.apache.mahout.cf.taste except what's in .hadoop . This is the non-distributed, non-Hadoop code. You can create a Recommender using this code, and have the framework wrap it in a ready-to-deploy .war file which exposes the recommender as a web service, via JWS.

This is the key documentation: https://cwiki.apache.org/confluence/display/MAHOUT/Recommender+Documentation

You might also be interested in the book Mahout in Action .

Answer 2

Mahout runs on Hadoop (according to the documentation) and if you google out there, there are projects that have used a rails stack to communicate with a hadoop cluster. But the basic process would be to running you hadoop cluster on a set of ec2 instances or some other clustering infrastructure and then interface that with you app through a database adapter. it looks like one that I have had some success with on other databases (oracle, mySQL, LDAP, jdbc) is datamapper and they are interested in developing a hadoop adapter: http://wiki.github.com/datamapper/dm-core/ruby-summer-of-code-ideas

this is probably the most promising library i saw for your purposes though: http://mrflip.github.com/wukong/INSTALL.html

Answer 3

Go ahead with JRuby and implement the jar files provided with Mahout binary distribution. You can seamlessly integate it.