Multi-attribute learning mechanism for network control and management

Presents a learning mechanism to extract an action pattern under various network conditions. A multi-attribute learning mechanism has the ability to indicate the success probability of each action pattern under given network conditions. Not only observed network information but also qualitative factors can be used in this mechanism. It is an effective way to support decision-making to take multiple factors including probabilistic or uncertain factors. A dynamic routing scheme employing this mechanism and a performance evaluation are shown as an application example.