An XCS-based approach to classification learning on continuous-valued attributes

It is a challenge for XCS to handle the problems presented by large continuous-valued search spaces. This paper proposes two models for handling continuous-valued attributes with the aim of improving XCS performance based on learning speed and ability to adapt the different test environments. Re su lt s show that our proposed hybrid XCS-based learning system provides the comparable classification accuracy and generates a complete set of simple rules that some kind of learning classifier cannot induce.