Comparison Between Classical and Advanced Naïve Bayes Classifier

Abstract

Due to the quite number of applications which the main problem of it is to classify new observations into known groups, several supervised classification methods were set to solve such problems depending on a dataset used to build a classification function, one of the most important methods is Naïve Bayes Classifier NB.

This research has addressed the basic definitions related to supervised classification and naïve bayes classifier, which it defines how the classifier works with related ideas, furthermore, it introduces the advanced status of NB, which is called Flexible Bayes FNB.

A practical comparison has been made between the basic NB and advanced FNB status using datasets of different applications in terms of size and dimensions, the results showed that FNB perform better than NB in most applications, also, it reveals the importance of probability distributions of the independent variables and the size of training data to achieve a higher accuracy and to show the difference of accuracies between NB and FNB, and which one is preferred to use.