It’s best to run this JUnit test from either IntelliJ, http://www.jetbrains.com/idea/download/, or Eclipse, http://www.eclipse.org/downloads/. Create a Java project, create a package taste and a class TanimotoDemo. Now add the mahout jars as dependencies to your project and you can run TanimotoDemo as a JUnit test.

It’s possible to run it from the command line with

java -cp all:your:dependencies:here org.junit.runner.JunitCore package.to.TanimotoDemo

but you need to setup your classpath and an IDE makes that a lot easier.

Note btw that this a non-distributed recommendation, it does not run on Hadoop. I will dive into this subject in the future and blog about it.

My mahout trunk is in /Users/nkodner/development/mahout/trunk/

I’m not sure how to set up my classpath. I’m also not sure where to save TanimotoDemo.java.

Any help is greatly appreciated! I’m looking to try Mahout, specifically TanimotoSimilarity on top of my Hadoop cluster to see how it performs against my Python implementation.

There is an active and growing community behind it, so I’d say give it a try.

For a real-life example of Mahout, look into Mippin, a search engine with recommendation features. Check out http://blog.mippin.com/2009/09/may-we-recommend.html

My company is developing a recommendation engine. Do you recommend using Mahout as our core or are there better open source collaborative filtering systems out there? Also, do you know of any websites that use Mahout to provide recommendations? I would like to see it in action.

org.apache.mahout.cf.taste.common.NoSuchUserException

To clarify I think you should change the following:

userIdMap.put(0, customerAPrefs);
userIdMap.put(1, customerBPrefs);
userIdMap.put(2, customerCPrefs);

into:

userIdMap.put(CUSTOMER_A, customerAPrefs);
userIdMap.put(CUSTOMER_B, customerBPrefs);
userIdMap.put(CUSTOMER_C, customerCPrefs);