A Study of Classification Based on Bayes Classifiers

When solving the classification problems, it is common to apply feature selection as a pre-processing technique. In this paper, we do experiments to compare the abilities of some feature selection methods such as chi squared, symmetrical uncertainty and RelifF. Also, the performances of some classifiers in different datasets are compared. Results on different datasets show the Bayesian classifiers perform well, especially for hidden naive Bayes which is better than others. Also, the performance of symmetrical uncertainty for selecting relevant metrics is promising.