Is Always a Hybrid Recommender System Preferable To Single Techniques?
Himan Abdollahpouri, Adel Torkaman Rahmani, Alireza Abdollahpouri · International Journal of Computer Applications · 2013
Collaborative filtering (CF) recommender systems are typically unable to generate adequate recommendations in sparse datasets.Empirical evidence suggests that incorporation of a trust network among the users of a recommender system can significantly help to alleviate this problem.For this reason, some studies have been done on combining CF with trust-enhanced recommender system.In this study, we analyze the switching hybrid recommender system with the CF and trust-enhanced recommender system components from both rating coverage and mean absolute error point of view.Experiments on a dataset from Epinions.comprove that, although the rating coverage of this hybrid method is better than both (CF and trust-enhanced RS), but has lower accuracy than just using trust-enhanced RS.In other words, trust-enhanced RS outperforms the hybrid recommender system consisting of CF and trust-enhanced RS.Finally, we justify this result using analytical method.