Using accuracy-based learning classifier systems for imbalance datasets

XCS is one of the most powerful learning classifier systems. It combines reinforcement learning and genetic algorithm to create a set of rules representing the extracted knowledge from dataset. The main advantage of this system is to provide rule-based models that represent human-readable patterns. However, not too much public have yet been studied in imbalance dataset. In this paper, we propose a novel technique to develop XCS deal with imbalance dataset. The proposed technique uses adaptive perception rate for each rule to provide balance learning between major and minor class. The experiment show that the propose technique can classify all level of imbalance classes on the well-know Boolean logic benchmark task - multiplexer problem.