Improving Recommender Systems by Incorporating Similarity, Trust and Reputation

Recommender systems using traditional collaborative filtering suffer from some significant weaknesses, such as data sparseness and scalability. In this study, we propose a method that can improve the recommender systems by combining similarity, trust and reputation. We modify the way that neighbors are selected by introducing the trust and reputation metrics in order to develop new relations between users so that it can increase the connectivity and alleviate the data sparseness problem. Throughout our 2 different scenarios of experiment simulations conducted on MovieLens dataset and the comparison of our results with other trust-based collaborative filtering research, we found out that our proposed method outperforms for better recommendations in an effective way.