Guided rule discovery in XCS for high-dimensional classification problems
AI'11 Proceedings of the 24th international conference on Advances in Artificial Intelligence
| Hi-index | 0.00 |
XCS is a learning classifier system that combines a reinforcement learning scheme with evolutionary algorithms to evolve rule sets on-line by means of the interaction with an environment. Usually, research conducted on XCS has mainly focused on the analysis and improvement of the reinforcement learning component, overlooking the evolutionary discovery process to some extent. Recently, the first efforts towards analyzing and designing new operators for the evolutionary algorithm have been done. The selection pressure produced by different selection schemes has been studied and the rule representation of XCS has been extended to adapt evolution strategies as the discovery component of the system. This paper continues on the analysis of the evolutionary algorithms in the on-line architecture by analyzing the role of the crossover operator in the original XCS and XCS based on evolution strategies. A new recombination operator, inspired by the BLX crossover operator in realcoded genetic algorithms, is designed for XCS. The new recombination operator is experimentally compared with the traditional crossover operator of XCS on a collection of real-life classification problems. The results show the competence of the new operator, providing the best results, on average, on the tested domains.