Abstract: Recently online social networks have become the major platform with millions of registered users on the Web．The amount of information is increasing so quickly that users can’t handle the information overload without the support of recommendation methods．Traditional recommendation methods have a limited performance in the context of social recommendation due to not considering the social network information,such as trust．Trust-based methods attempt to introduce a trust metric during the social recommendation．However,most of these methods are based on the explicit trust statements expressed by users,which are not available in the social networks such as Facebook,Twitter and Sina Weibo．This paper presented a trust metric to quantitatively measure the recommendation trust between pairs of users by aggregating the implicit trust and trust propagation values．We proposed a trust-based latent factor model,which incorporates the pairwise recommendation trust values into the probabilistic model for top-k item recommendation．The experiments on Sina Weibo demonstrate that our method outperforms the traditional recommendation methods and trust-based methods.

Key words: Social trust computation,Latent factor model,Recommender system,Social recommendation,Social network