Two-level feature selection for naive bayes with kernel density estimation in question classification based on Bloom's cognitive levels

Abstract

This paper proposes a two-level feature selection to improves Naïve Bayes with kernel density estimation. The performance of the proposed feature selection is evaluated on question item set based on Bloom's cognitive levels. This two-level feature selection contains of filter and wrapper based feature selection. This paper uses chi square and information gain as the filter based feature selection and forward feature selection and backward feature elimination as the wrapper based feature selection. The result shows that the two-level feature selection improves the Naïve Bayes with kernel density estimation. The combination of chi square and backward feature elimination give more optimal quality than the other combination.