With the rapid growth of information on the World Wide Web, social recommendations have appeared as one of the most important roles attracting growing attentions from researchers. Social recommendations enable a form of efficient knowledge for users and help them share contents with others. There are many studies in this area focusing on using trust-relationships in recommendation algorithm, which has become a major trend in recommendation algorithms that used for searching information precisely, feasibly and efficiently, but they neglect how to build the trust-relationships framework at start. In this work, an algorithm, called
, is proposed for building a trust-relationship framework to improve the social recommendations when there is no or too few of available trust-relationships. Here, we first construct a graphical model based on a binary-type vertex relationship, where discusses the explicit and potential connections among users and recommended items. On this basis, we implement a common-used collaborative filtering recommendation algorithm to deal with the situation of enough available trust-relationships existing, and then present
, which builds trust-relationship framework as a supplement. Finally, we crawl through data from three famous recommender websites, i.e., del.icio.us, Myspace and MovieLens and use them in experiments to show that
can suggest relevant items to users’ tastes and perform better than collaborative filtering algorithm in precision and stability.