Optimization Feature Selection for Classification of Manuscript Grouping

Abstract

Currently, data mining can be very useful to data computational, retrieve important and relevant information. Classification is a popular data mining function. The limitation of the classification model must use a feature selection to investigate with the real system. This paper, we used the Information Gain approach, which is the one method of feature selection, to improve performance Naïve Bayes and Decision Tree model for a manuscript grouping. The case study used 177 manuscripts from National Science Conference 4^th which has 4 groups including mathematic healthcare computer science and other related fields. The results shown the Naïve Bayes model with information gain weight 0.03 has the best accuracy at 84% and 444 features. Moreover, the Decision Tree model with information gain weight 0.20 has accuracy only 60% and 9 features. Concluding, the Naïve Bayes classification method with the information gain method can apply the best classification model with this case study.