Evolutionary rule-based systems for imbalanced data sets

This paper investigates the capabilities of evolutionary online rule-based systems, also called Learning Classifier Systems (LCSs), for extracting knowledge from imbalanced data. While some learners may suffer from class imbalances and instances sparsely distributed around the feature space, we show that LCSs are flexible methods that can be adapted to detect such cases and find suitable models. Results on artificial datasets specifically designed for testing the capabilities of LCSs in imbalanced data show that LCSs are able to extract knowledge from highly imbalanced datasets. When LCSs are faced with real-world problems, they demonstrate to be one of the most robust methods compared with instance-based learners, decision trees and support vector machines. Moreover, all the learners benefit from resampling techniques. Although there is not a resampling technique that performs best in all datasets and for all learners, those based in oversampling seem to perform better in average. The paper adapts and analyses LCSs for challenging imbalanced datasets and sets the bases for further studying the combination of resampling techniques plus learner best suited to a specific kind of problem.