Learning approach for dealing with unfair ratings in service provision

Service-oriented environments are open, dynamic and highly distributed environments with various autonomous agents. Finding the desirable services among others is a major challenge for users, who can optimize its performance by utilizing services with good qualities. This challenge is sometimes addressed by relying on recommends about qualities of services collected from other agents in the environment. However, there is no guarantee that all agents give fair recommends about all services and their Qualities. The presence of deceptive agents makes it necessary to develop methods for distinguishing fair and unfair agents from each other and providing users with reliable trust information. Generally, a service can be characterized by a number of attributes and each attribute has different importance for different users. Many trust models have been developed based on feedback mechanisms, but most of them are unable to detect inaccurate information that a deceptive agent may generate, and can barely separate fair and unfair agents. In the proposed method we use learning automata to separate fair and deceptive agents based on the feedback from the recommends they have given on different services and service attributes in the environment. The experimental simulations have shown that the new method outperforms other existing models.