Learning Pattern Tree Classifiers Using a Co-Evolutionary Algorithm

Pattern tree induction has recently been introduced as a novel method for classification. Roughly speaking, a pattern tree is a hierarchical, tree-like structure, whose inner nodes are marked with generalized (fuzzy) logical operators, and a pattern tree classifier consists of one such tree per class. Since a pattern tree can thus be considered as a kind of logical characterization of a class, the approach is very appealing from an interpretation point of view. Yet, as will be argued in this paper, the method that has originally been proposed for learning pattern trees is not optimal and offers scope for improvement. To overcome its disadvantages, we propose a new method which is based on the use of co-evolutionary algorithms. Experimentally, it will be shown that our approach is indeed able to outperform the original learning method in terms of predictive accuracy.