Using case-based reasoning as a reinforcement learning framework for optimisation with changing criteria.