Resumo

PEREIRA-TOLEDO, Alain; LOPEZ-CABRERA, José D.  e  QUINTERO-DOMINGUEZ, Luis A.. Experimental Study for the Comparison of Naïve Bayes with other Bayesian Classifiers. Rev cuba cienc informat [online]. 2017, vol.11, n.4, pp.67-84. ISSN 2227-1899.

An experimental study is conducted for the comparison of Naïve Bayes classification performance with other Bayesian classifiers. There are some other similar reports of experiments, but they do not use the more suitable methodology proposed by Demšar. That is why the purpose is revisiting Naïve Bayes in comparison with other Bayesian classifiers. In consequence, we chose five Bayesian classifiers and Naïve Bayes; all of them are implemented in WEKA. In addition, we chose three set of databases for the three set of experiments: a general experiment, an irrelevant feature experiment and a noisy values experiment. Finally, we showed by means of the Demšar methodology that there is empirical evidence for stating that Naïve Bayes still is a competitive alternative to the rest of the selected Bayesian classifiers.

Palavras-chave : Naïve Bayes; Bayesian classifiers; experimental study; irrelevant features; noisy values.

        · resumo em Espanhol     · texto em Espanhol     · Espanhol ( pdf )