An Improved Continuous-Action Extended Classifier Systems for Function Approximation

Due to their structural simplicity and superior generalization capability, Extended Classifier Systems (XCSs) are gaining popularity within the Artificial Intelligence community. In this study an improved XCS with continuous actions is introduced for function approximation purposes. The proposed XCSF uses “prediction zones,” rather than distinct “prediction values,” to enable multi-member match sets that would allow multiple rules to be evaluated per training step. It is shown that this would accelerate the training procedure and reduce the computational cost associated with the training phase. The improved XCSF is also shown to produce more accurate rules than the classical classifier system when it comes to approximating complex nonlinear functions. © 2015 The Authors. Published by Elsevier B.V. Peer-review under responsibility of scientific committee of Missouri University of Science and Technology.