Fuzzy Classification with Multi-objective Evolutionary Algorithms

In this work we propose, on the one hand, a multi-objective constrained optimization model to obtain fuzzy models for classification considering criteria of accuracy and interpretability. On the other hand, we propose an evolutionary multi-objective approach for fuzzy classification from data with real and discrete attributes. The multi-objective evolutionary approach has been evaluated by means of three different evolutionary schemes: Preselection with niches, NSGA-II and ENORA. The results have been compared in terms of effectiveness by means of statistical techniques using the well-known standard Iris data set.