Research On Trust Based Recommendation Method And Its Application

During the past decade, in order to solve the problems which the traditional recommender system faced and to improve the quality of recommendation, many researchers added trust to the recommending process, studying how to improve the recommendation algorithm using trust. Such studies are called trust-based recommendation algorithm (system), which have been shown to a good solution to the problems the traditional recommendation system faced. The trust based recommendation method is a kind of social recommendation, for it making use of trust information from social networks. In this paper, we study how to make use of trust to make recommendations, what kind of information should be used as a source of trust information, how to specify and quantify the abstract concept of trust (the trust model or the trust computing method) and how to add the quantified trust to the recommending process.In order to improve the quality of recommendation using trust, for the general E-commerce recommendation, we propose a novel trust-based recommendation method using the properties of trust network. The trust network is a social network where nodes are inter-linked by their trust relations. However, little is known about the structural properties of trust network, and little exploration of using these properties to develop the trust model has been done. Firstly, we put forward a method leveraging implicit trust information to develop a trust network. Secondly, we illustrate that this trust network is a dynamic small-world network by analyzing its properties. Finally, we propose a novel trust-based recommendation method using the properties of trust network. Experimental results clear showed that:our proposed method is superior to the classic collaborative filtering on the accuracy performance.In such a customer-relationship-based microblog platform, the interactions between users are very important. Microblog user recommendation helps users find microblog which they may be interested in, meeting their demands of information acquiring, sharing and propagating on the microblog. Applying the above trust-based recommendation method to microblog user recommendation problem, we propose trust-based microblog user recommendation methods. Firstly, we propose a method to build users’ rating to others using available information from microblog and the user trust computing method (the user trust model). By analyzing the available information from microblog, we chose such information that can reflect the trust or the interest and identity of a user to another one. We build users’ rating and trust to others using these information. Secondly, on this basis, we put forward two microblog user recommendation methods using trust. One is the direct improvement of the FoF user recommendation method, and another is a trust-based microblog user recommendation method using the idea of trust-based collaborative filtering. Finally, we evaluate the quality of our user recommendation methods using the experimental data collected from Sina Weibo. The results demonstrate the effectiveness of these methods.