FS-XCS vs. GRD-XCS: An analysis using high-dimensional DNA microarray gene expression data sets

XCS, a Genetic Based Machine Learning model that combines reinforcement learning with evolutionary algorithms to evolve a population of classifiers in the form of condition-action rules, has been used successfully for many classification tasks. However, like many other machine learning algorithms, XCS becomes less effective when it is applied to high-dimensional data sets. In this paper, we present an analysis of two XCS extensions – FS-XCS and GRD-XCS – in an attempt to overcome the dimensionality issue. FS-XCS is a standard combination of a feature selection method and XCS. As for GRD-XCS, we use feature quality information to bias the evolutionary operators without removing any features from the data sets. Comprehensive numerical simulation experiments show that both approaches can effectively enhance the learning performance of XCS. While GRD-XCS has obtained significantly more accurate classification results than FS-XCS, the latter has produced much quicker execution times than the former.