Introduction

This plugin provides recommendations based on machine learning library Apache Mahout.

Configuration

This plugin can defined multiple recommenders.

Recommenders should load a large set of data to be able to provide recommendations (learning phase). Each set of data should provide those information:

• The user ID (as long)
• The item ID (as long)
• A preference value (as float)

Recommenders should be listed into the recommendation.properties file and for each one the dataset source should be defined (table and columns) as below :

#recommendation.properties

# List of recommenders    
recommendation.recommendersList=rec1,rec2

# recommender 1
recommendation.recommender.rec1.dataSource=portal
recommendation.recommender.rec1.preferenceTable=recommendation_default
recommendation.recommender.rec1.userIDColumn=id_user
recommendation.recommender.rec1.itemIDColumn=id_item
recommendation.recommender.rec1.preferenceColumn=preference_value
 
# recommender 2  
recommendation.recommender.rec2.dataSource=portal
...    
                

Usage

The recommendations can be provided in Java by the RecommendationService

List<RecommendedItem> list = RecommendationService.instance().getRecommendations( strRecommender, lUserId, nCount );
                

Or through HTTP

http://myserver.com/servlet/plugins/recommendation/?id_user=2&count=2&recommender=test
                

Default values for the recommender and the count parameters can be defined into the recommendation.properties file, so the url can be limited to :

http://myserver.com/servlet/plugins/recommendation/?id_user=2