Service Reliability Prediction Method for Service-Oriented Multi-agent System

With gaining the benefit of the cross-language and cross-platform characteristics of Service-Oriented Computing (SOC), Service-Oriented Multi-Agent System (SOMAS) has attracted many researcher´s attention. SOMAS gives intelligent and autonomous properties to overcome the limitations of SOC such as hardly adaptive to the open and dynamic environment. In this paper, we focus on the service reliability prediction of SOMAS. The service reliability is a critical quality attribute to measure the degree of successful service operation. However, the research on the service reliability prediction of SOMAS has not been studied a lot because this is relatively a new concept and can be affected many dynamic context factors such as workload-level of devices executing the agents. We propose a reliability prediction framework using matrix factorization with explicit factor analysis. It employs workload of the agent and network sensitivity of the task to reflect the characteristics of SOMAS in prediction method. Through the simulation, proposed approach shows the best prediction performance compared to the state-of-the-art collaborative filtering techniques.