Abstract
In a multi-agent recommender system, the knowledge of an agent in a given domain and the trust attributed to it relative to this domain are two main criteria for a better recommendation. On this basis, we proposed a new multi-agent recommendation approach that exploits both the knowledge and the trust of an agent to determine the expertise degree of an agent in a given domain. Our approach consists of two phases: a modeling phase using the Fuzzy Formal Concepts Analysis (FFCA) and a decision-making phase using the theory of belief functions.