FALCS AND ITS APPLICATIONS IN CONTROL AND DECISION-SUPPORT

Learning classifier systems are machine learning algorithms based on evolutionary computing, and utilizing components of both reinforcement learning and genetic algorithms. Providing an input-output mapping, they rely on populations of rules that are modified based on their utility,for the given task at hand through a series of predefined procedures. As such, they have been successfully applied in many different applications; nevertheless, continued research is being performed to improve their performance, to expand their applicability to new problems and to gleam better understanding of their theoretical properties. This paper describes research endeavours to extend conventional learning classifier systems with the concepts of fuzzy logic, so they can better handle problems with real-valued rule representations. The paper also illustrates several possible applications, for which the proposed system could be used.