Case-based reasoning in agent-based decision support system

Abstract

The paper describes decision support system for modeling, control, and simulation of continuous, as well as discrete event systems. Models, control methods, and tools are in database specified by their attributes. Each attribute's weight is initially estimated according to importance and classification power of a given feature. Automatic learning of attributes weights uses the answers of the users after simulation provided by the system to increase the quality of case-based reasoning. The proposed learning algorithm guarantees convergence of the attributes weights to relatively steady values yielding to Case-based reasoning of best quality. If simulation of a new case was successful from the point of view of the user, the new case is added to case base.