Issue No.01 - January (2008 vol.20)
UCS, a sUpervised Classifier System, is an accuracy-based evolutionary learning classifier system for data mining classification tasks. UCS works through two stages: exploration and exploitation. During the exploration phase, a population of rules is evolved in order to represent a complete solution of the target problem. The exploitation phase is responsible for applying the rule-based knowledge obtained in the first phase when the system is exposed to unseen data. The representation of a rule in UCS as a univariate classification rule can be easily seen in a symbolic form, which is easy for a human to understand and comprehend (i.e. expressive power). However, the system may generate a large number of rules to cover the input space. Artificial neural networks normally provide a more compact and accurate representation. However, it is not a straightforward task to understand the network. In this paper, we propose a novel way to incorporate neural networks into UCS. The approach offers a good compromise between compactness, expressiveness, and accuracy. By using a simple artificial neural network as the classifier's action, we obtain smaller/compact population size, better generalization, while maintaining a reasonable level of expressiveness. We also apply negative correlation learning (NCL) during the training of the resultant neural network ensemble. NCL is shown to improve the generalization of the ensemble.
Rule-based processing, Representations (procedural and rule-based), Learning, Knowledge modeling
Hai H. Dam, Hussein A. Abbass, Chris Lokan, Xin Yao, "Neural-Based Learning Classifier Systems", IEEE Transactions on Knowledge & Data Engineering, vol.20, no. 1, pp. 26-39, January 2008, doi:10.1109/TKDE.2007.190671