Abstract
The cold start problem is a potential issue in computer-based information systems that involve a degree of automated data modeling. Specifically, the system cannot infer a rating for users or items that are new to the recommender system when no sufficient information has been gathered. Currently, more websites are providing the relationships between users, e.g., the trust relationships, to help us alleviate the cold start problem. In this paper, we proposed a trust-based recommender model (RSOL) that is able to recognize the user's recommendation quality for different items. A user's recommendation quality contains two parts: "Rating Confidence" - an indicator of the user's reliability when rating an item, and "Proximity Prestige" - an indicator of the user's influence on a trust network. In our experimental results, the proposed method outperforms the Collaborative Filtering and trust-based methods on the Epinions dataset. © 2013 Springer-Verlag.