The Abstract of Thesis Classifier by Using Naive Bayes Method

The thesis is a requirement for graduation from Bumigora university. The final year student’s problem is determining the research topic because the undergraduate thesis collection of Computer Science is not grouped or classified based on student competencies. The purpose of this study was to compare the performance of the naïve Bayes method with TF-IDF weighting and without TF-IDF weighting for the classification of thesis topics based on the abstract. The stages of this research are data collection, text pre-processing, term weighting with TF-IDF and without TF-IDF, Naïve Bayes method implementation, and result evaluation. Based on the results of the tests that have been done, the naïve Bayes method with TF-IDF has an accuracy of 81.74%, a precision of 86.1%, and a sensitivity of 80.15%. While the naïve Bayes method without TF-IDF weighting produces 88.69% accuracy, 89.76% precision, and 90.49% sensitivity. Thus, the naïve Bayes method without TF-IDF weighting has better performance than TF-IDF weighting for the classification of thesis topics based on the abstract.