Can Evolutionary Computation Handle Large Dataset?

Evolutionary Learning Classifier Systems (ELCS) were introduced by Holland a few decades ago. Since their birth, they were successfully applied to various data analysis domains. XCS is currently considered as state of the art ELCS. Earlier work have experimented with XCS on artificial problems or small datasets, and shown good results. However, XCS has not been tested on large datasets, particularly in the intrusion detection domain. This work investigates the performance of XCS on the 1999 KDD Cup intrusion detection dataset, a real world dataset approximately five million records, more than 40 fields and multiple classes with non-uniform distribution. Intuitively XCS fits well into this domain due to its online learning mode and its ability to evolve accurate and maximum general rules. We report the findings on experiments with the original XCS that show low accuracy. We then propose several modifications to XCS to improve its detection accuracy. The modified XCS improves remarkably on three of the four classes. The overall accuracy is equivalent to that of traditional machine learning algorithms, with the additional advantages of being evolutionary and online learner.