Multinomial Naïve Bayes Optimization with Information Gain for Library Book Classification

This research explores optimization in book classification activities at Politeknik Negeri Indramayu Library. The method commonly used by librarians to classify books is Dewey Decimal Classification (DDC). The DDC method allows librarians to classify book universally and systematically. However, it takes more effort and time to obtain a book classification label. This is not efficient, considering a large number of books in the library. For this reason, we propose an automatic book classification model using the text mining method. Based on the previous research, book classification model using the Multinomial Naïve Bayes (MNB) method has been conducted. The results of these studies indicate an accuracy value of 65.4%. However, the accuracy value still depends on the number of features of the dataset, so that the greater the number of features, the smaller the accuracy value. In this study, Information Gain (IG) method is proposed to select the features of a dataset in the pre- processing stage. MNB accuracy measurements are carried out based on before and after feature selection. 10-fold cross-validation is used to validate the classification model. The results showed an increase in the accuracy of MNB by 6.6%