A Hybrid Recommendation System Using Trust Scores in a Social Network

Various techniques for personalized recommendation have been studied. One of the most promising such techniques is collaborative filtering (CF). However, CF is unable to produce high quality recommendations when user rating data is lacking or insufficient. To address this “sparsity” problem of CF systems, this paper proposes a hybrid recommendation system. The proposed system improves recommendation quality by exploiting trust scores between users in a social network. The proposed system overcomes the weakness of CF for sparse user ratings databases and yields better performance than the conventional CF method.