Abstract: In ubictuitous computing environment, services recommendation needs to satisfy the system's mobility and heterogeneity. A multi Agent service recommendation mechanism was put forward, bayesian network and clustering method and update mechanism were used. It is fully taken into account that contexts affect the result of recommendation. The system consists of Agents which accomplish the functions such as history context collecting, knowledge training, poficymaking recommendation and self-learning. Furthermore, by designing the information of communication, flow control and data shared-path among Agents were constructed.

Key words: Ubiquitous computing, Service recommendation, Multi Agent, I3ayesian network