Trust and Reputation Based Model Selection Mechanism for Decision-Making

In open and distributed environment, model selection of decision-making is difficult due to the lack of model´s knowledge. In order to find an effective mechanism for decision parties, semantics based trust and reputation computations are introduced into decision-making. Trust between decision parties and models are divided into trust dependence and trust relationship which demonstrate the matching degree between the semantics from sponsor model requirements and the semantics of candidate model´s capacities, and the subjective reliability between sponsor party and provider party respectively. Then, a reputation ranking method is proposed to build up a neutral and whole belief about a decision party. Reputation ranking is composed by three facets: time limitation, source identity and ranking delay. According to the trust and reputation computation, we present a trust and reputation based model selection mechanism for decision party. Finally, the experiment shows that our approach is feasible and effective.