MÉTODOS DE COMBINAÇÃO DE CLASSIFICADORES

Abstract

The methods of combining classifiers are considered a problem in many application areas. With the objective knowledge of standards and a systematic investigation. Pattern recognition is based on objects in order to solve real life problems. There are two types of pattern recognition: supervised and unsupervised. The classifiers are divided into: Bayesian, k-NN classifiers away and neural networks. The combination of multiple classifiers, we propose the development of reliable recognition system, as well as better performance of individual classifiers. The goal of combining classifiers is to improve the efficiency of decision making, adopting rules of combination of multiple stages (the product rule, sum rule, min rule and max rule) the combination of several independent classifiers is a general problem that occurs in various application areas of pattern recognition.

Index Terms - application, classifiers, combining, decision.