Learner classifier systems: Theory and applications - A state-of-the-art review

Learning Classifier Systems (LCSs) are rule based machine learning systems that use genetic algorithms as their primary rule discovery mechanism. With in these systems, a 'population' of condition/action rules is evolved to fi nd those that cope best with the system's external environment. This idea a learning system based on internal variation and selection - contrasts with more widely studied learning systems based on artificial neural networks, and has found many potential applications in process control, robots contro l, telecommunication and etc. As such, there has been extensive research in this field. This review attempts to collect, organize, and present in;a un ified way some of the most representative publications on learning classifi er systems. To organize the literature, the paper presents a categorization of the main components of LCSs, which include rule set and message system, performance system, reinforcement algorithm and rule discovery system and presents some recent advancements. Also, the paper describes some of the mo st significant applications of LCSs.