Spherical Conditions with XCSc

This paper presents a novel approach to clustering on the spherical condition structures to accuracy-based Learning Classifier System (sXCSc). Our approach achieves this by exploiting evolutionary computing, reinforcement learning and the generalization mechanisms inherent to XCS. The purpose of our work is to develop learning rules which accurately describe clusters without prior assumptions as to their number within a given dataset. Experiments on synthetic and real datasets confirm that improvements in both accuracy and finding number of clusters are achieved.