Aggregation of fuzzy classifiers using coupled map lattices

Abstract

This paper proposes a technique for aggregating a group of fuzzy classifiers using a coupled map lattice. First a the training data set is divided into several disjoint groups. Each group is used for training a classifier using a fuzzy classification technique. Then, each fuzzy classifier is associated to one site in a coupled map lattice. In order to predict the class of a given data sample, the sample is presented to each fuzzy classifier and the prediction is evolved using the dynamics properties of the coupled map lattice. The final prediction is the fuzzy voting of the classifiers after evolving them. The proposed approach is tested with several toy and real data sets in order to determine its performance. © 2006 IEEE.