Learning Fuzzy Classifiers with Evolutionary Algorithms Mauro L. Beretta 1 and Andrea G. B. Tettamanzi 2 1 Genetica S.r.l. Via S. Dionigi 15, I-20139 Milano, Italy E-mail:beretta@genetica-soft.com 2 Università degli Studi di Milano Dipartimento di Tecnologie dell'Informazione Via Bramante 65, I-26013 Crema (CR), Italy E-mail: andrea.tettamanzi@unimi.it Abstract. This paper illustrated an evolutionary algorithm which learns classifiers, represented as sets of fuzzy rules, from a data set containing past experimental ob- servations of a phenomenon. The approach is applied to a benchmark dataset made available by the machine learning community. 1 Introduction Classification, or pattern recognition, is the task of assigning an instance to one of two or more pre-defined classes, or patterns based on its characteristics. A classi- fier is an object which computes a mapping from instance data to their (hopefully correct) class. An example of classification is, given a printed character (the in- stance), to associate it with the letter of the alphabet (the pre-defined class, or the pattern) it represents. Known patterns are usually represented as class structures, where each class structure is described by a number of features. These features define a feature space, wherein data are defined. In this paper we illustrate an approach to automatically learning the fuzzy deci- sion rules of a classifier by evolutionary algorithms. 2 The Problem Classification can be formulated as an optimization task by defining an objective function of the form Z=F(classifier; data set) (1) which measures the accuracy of a classifier when applied to the data set. Let us assume that a data set consists of N records, each one with a class attrib- ute c i ∈ {0, M - 1} where M is the number of classes. Let c * indicate the classi-