Guided Rule Discovery in XCS for High-Dimensional Classification Problems

XCS is a learning classifier system that combines a reinforcement learning scheme with evolutionary algorithms to evolve a population of classifiers in the form of condition-action rules. In this paper, we investigate the effectiveness of XCS in high-dimensional classification problems where the number of features greatly exceeds the number of data instances – common characteristics of microarray gene expression classification tasks. We introduce a new guided rule discovery mechanisms for XCS, inspired by feature selection techniques commonly used in machine learning. The extracted feature quality information is used to bias the evolutionary operators. The performance of the proposed model is compared with the standard XCS model and a number of well-known machine learning algorithms using benchmark binary classification tasks and gene expression data sets. Experimental results suggests that the guided rule discovery mechanism is computationally efficient, and promotes the evolution of more accurate solutions. The proposed model performs significantly better than comparative algorithms when tackling high-dimensional classification problems.