Using a Different mahout Algorithm from the taste-getting-started Example

We were successful in running the taste-getting-started example as seen here: http://teachitshem.blogspot.com/2011/04/simple-java-wicket-using-mahout.html. However, mahout offers other recommendation algorithms which are so interesting to study. For instance, the PearsonCorrelationSimilarity whose parameters are similar to that EuclideanDistanceSimilarity used in the original taste-getting-started example.

How then can do this?

First, we duplicate the taste-getting-started folder say to tgs and we achieve this by issuing cp taste-getting-started tgs –R in the command line.

Second, we change the node <artifactId>taste-getting-started</artifactId> to <artifactId>tgs</artifactId> from the pom.xml file.

Third, open the recommeder-context.xml file at src/main/webapp/WEB-INF folder and replace euclideanDistanceRecommeder with pearsonCorrelationSimilarityRecommender, euclideanDistanceSimilarity with pearsonDistanceSimilarity, and similarity.EuclideanDistanceSimilarity with similarity.PearsonCorrelationSimilarity.

Fourth, run mvn package against tgs.

Fifth, in order to compare results with the original taste-getting-started we have run both the original taste-getting-started and the new tgs web applications using different ports. Thus for taste-getting-started we can use the usual mvn jetty:run-war that uses port 8080 but for tgs we must use something like mvn –Djetty.port=9090 jetty:run-war which uses the URL http://localhost:9090/tgs/movies.

Simple eh?

Simple Java Wicket using Mahout

While continuing the study of mahout, it is inevitable to write code that uses the various algorithms stored with it. So I asked a colleague to help out and one option was to start with a java wicket found here: http://blog.jteam.nl/wp-content/uploads/2010/04/taste-getting-started.zip

Using it is quite straightforward. Download the zip file. Unzip it to some folder (say taste-getting-started in the mahout machine). Using Ubuntu’s Terminal application, go to that folder and run mvn package. After packaging is complete, run mvn jetty:run-war. Once the jetty server is up, you can make use of the wicket by browsing http://localhost:8080/taste-getting-started/movies.

Although the above is the straightforward usage of the said wicket, setting it up requires some effort such as

• downloading the MovieLens’ 100K Ratings Data Set: http://grouplens.org/system/files/ml-data_0.zip
• Unzipping it to some other folder.
• Copying u.data to taste-getting-started/src/main/resources/grouplens/100K/ratings
• Copying u.item to taste-getting-started/src/main/resources/grouplens/100K/data
• Executing the initialize_movielens_db.sql using the following command mysql --user=mysqluser --password=mysqlpassword < initialize_movielens_db.sql while at the taste-getting-started/src/main/resources/sql folder.

Oh and if you do not have MySQL yet, install it by issuing the following command sudo apt-get install mysql-server mysql-client.