Convergence of agent reputation with Alpha-Beta filtering vs. a fuzzy system

The prediction of behaviours in dynamic and subjective environments is an interesting issue in the way to obtain a complete delegation of human-like decisions in autonomous agents. Complex mathematical algorithms have been proposed by academic researchers to be applied in agent-based recommendation systems. In this paper we propose to use a classic estimation method known as alpha-beta filtering. This innovative approach is compared with a previous proposal (called AFRAS) from the authors that it is based on fuzzy sets and that has been successfully compared with some of the most representative reputation models and with another classic estimation method (Kalman). This paper evaluates how consumer agents applying both methods predict the behaviour of provider agents focusing on the velocity of convergence of the predictions to the real behaviour where such behaviour adopts a prefixed variability