-
The user similarity and item similarity algorithms do not scale well. You need to precompute the correlations. The Duine Framework has support for storing the correlations in a database. Besides precomputation, you also have to make sure that the database access is cached.
Faster hardware will move the limit up, but in the end, with millions of users, you will need a better algorithm...
2009-08-05 13:08:48 UTC in The Duine Framework
-
The recommender currently uses rule based prediction strategies. It would be nice to extend the recommender with Bayesian-based or CBR-based strategies.
2009-02-26 13:55:48 UTC in The Duine Framework
-
Currently the recommender framework uses time-based testing. An alternative is to use 80/20 testing: Use 80% of the dataset (random selection) to learn from, then use the remaining 20% to validate the prediction performance. This could be even repeated with new 80% samplings for cross-validation.
2009-02-26 13:53:45 UTC in The Duine Framework
-
In the How to build http://www.duineframework.org/build.html a remark is made about the installation of the Sun jdbc stdext file in the maven repository. Now it states:
mvn install:install-file -Dfile=jdbc-stdext-2.0.jar -DgroupId=javax.sql -DartifactId=jdbc-stdext -Dversion=2.0 -Dpackaging=jar
This should be:
mvn install:install-file -Dfile=jdbc2_0-stdext.jar -DgroupId=javax.sql...
2009-02-26 13:11:39 UTC in The Duine Framework
-
jreitsma changed the public information on the The Duine Framework project.
2009-02-09 13:10:39 UTC in The Duine Framework
-
jreitsma changed the public information on the The Duine Framework project.
2009-02-09 13:09:54 UTC in The Duine Framework
-
jreitsma changed the public information on the The Duine Framework project.
2009-02-09 09:05:35 UTC in The Duine Framework
-
jreitsma changed the public information on the The Duine Framework project.
2009-02-05 16:06:41 UTC in The Duine Framework
