A skill refinement learning model for rule-based expert systems

Research in equipping rule-based expert systems with skill refinement behavior by utilizing the recognize-act control mechanism is described. An overview of expert system skill refinement is provided. A skill refinement model for generating plans is then presented. Two closely coupled and mutually supportive mechanisms characterize this model: a rule-selecting mechanism (corresponding to a buyer-selecting procedure) that dynamically incorporates the concept of multiple selection/preference criteria into the conflict resolution process, and an economics-based credit assignment mechanism (corresponding to a capital reallocation procedure) that uses an inference engine´s experiences to update the potentiality of each rule participating in the problem-solving process. A mathematical description of the model is given. An example is provided to illustrate the inference engine´s skill refinement and the applicability of the model is discussed.<>