Comparison of Supervised Learning Algorithm for Classification of Thesis Titles Based on Lecturer Fields

Abstract

At the level of education, especially for S1, the graduation requirement is to complete the thesis. In preparing the thesis, students are accompanied by a guidance lecturer who will direct and as a place to consult. The case is still there are students who are confused to take a thesis. There are several reasons that they do not have a title to be submitted, and are confused to choose a tutor who matches their title or theme. Sometimes on campus, students can get a mentor, but not in accordance with the field, and not in accordance with the theme of the thesis title. Therefore, in this study will make a classification of lecturer fields based on student titles. The data used as many as 1598 was taken from the campus of AMIKOM Yogyakarta University by adding some new data. With the lecturer in accordance with the field, it will be easier to guide students. This study conducted stages of labeling, text preprocessing, and word weighting or called TF-IDF (Term Frequency – Inverse Document Frequency). After that, the data split will be classified with naive bayes classifier (NBC), K-Nearest Neighbor (K-NN), and Support Vector Machine (SVM) algorithms. The performance of the three algorithms is compared to find out the performance of the algorithm is good. The results showed the Support Vector Machine (SVM) algorithm performed better by producing an accuracy of 89.24%, while the Naive Bayes Classifier (NBC) algorithm produced an accuracy of 88.29%, and the K-Nearest Neighbor (KNN) algorithm with a k value of 18 produced an accuracy of 85.14%.