Ensemble case based learning for multi-agent systems