Towards a Self-Organising Mechanism for Learning Adaptive Decision-Making Rules

Systems plunged into dynamic environments need evolving behaviours in order to self-adapt to these changes. These behaviours cannot be predetermined because it is impossible to list exhaustively all the situations the system may be faced with. Therefore, it becomes necessary to define real time algorithms that enable systems to autonomously adapt their behaviours to the current context. This paper focuses on behavioural rules learning. We propose, in that sense, a self-organisational approach based on local cooperative criteria that enable to discover triggering conditions of behavioural rules. Even if our approach intends to be generic, the principles and the evaluations have been defined in order to construct a system that enables the creation and the dynamic update of user profiles.