Novel implicit-trust-network-based recommendation methodology

Abstract

Recommender systems are known for helping e-commerce and entertainment sits to provide excellent services into their clients by finding those interests in a short time as possible. Recently, researchers have studied on trust relationships of people for improving user-based recommender systems. Hence, this study proposed an implicit approach for constructing a trust-network, which contains two parts. i. constructing an incipient trust-network by examining Similarity, Confidence and Identical Opinion of users; ii. Reconstructing the incipient trust-network as a telic trust-network by appraising its precision for predicting unknown-items rating. The study also presented methods for determining each of the similarity, the confidence, and the identical opinion. Furthermore, this study presented a recommendation strategy by merging two ways of ranking items based on their predicted rating and users' score to them, considering the trust-network. The study was evaluated on datasets of Ciao and FilmTrust for the trust-network correctness, the predicted-ratings error, and the quality of the recommendations. The results have shown, the scheme of this study has significantly improved the performance compared to some state-of-the-art techniques.