A rule based approach to group recommender systems

Abstract

The problem of building Recommender Systems has attracted considerable attention in recent years, but most recommender systems are designed for recommending items for individuals. In this paper we develop a content based group recommender system that can recommend TV shows to a group of users. We propose a method that uses decision list rule learner (DLRL) based on Ripper to learn the rule base from user viewing history and a method called RTL strategy based on social choice theory strategies to generate group ratings. We compare our learning algorithm with the existing C4.5 rule learner and the experimental results show that the performance of our rule learner is better in terms of literals learned (size of the rule set) and our rule learner takes time that is linear to the number of training examples. © 2011 Springer-Verlag.