GENERAL NAIVE BAYES STYLE FUZZY PROBABILISTIC CLASSIFIER

Abstract

Some kinds of naive Bayes style networks have been proposed, such as the multinomial naive Bayes, possibilistic naive Bayes and fuzzy Gaussian naive Bayes. However, a general formulation for a naive Bayes style fuzzy probabilistic network was not proposed yet. In this paper, we proposed a formulation for this kind of supervised classifier, using random variables without specifying any statistical distributions. This approach can be useful for classification purposes, when random variables can have different statistical distributions. A brief discussion about applications for data classification from health sciences is provided too.

Index Terms - Classification, Fuzzy Probability, Fuzzy Sets, Naive Bayes.