Categorization of Learning Materials Using Multilabel Classification

Adaptive learning can adjust learning materials based on students' individual abilities. To facilitate the selection of appropriate materials, the categorization of learning materials can be done first. This study aims to categorize learning materials based on topics and subtopics with multilabel classification. Multilabel problem is handled by problem transformation approach. The problem transformation methods used are Binary Relevance, Label Powerset, and Classifier Chain. While the classification algorithms are Naive Bayes, SVM, and Random Forest. The dataset used in this study is 448 learning materials which are science subject materials that include biology, physics, and chemistry for junior high school students. The evaluation results show that the best combination is achieved by Binary Relevance method and SVM algorithm with accuracy value of 0.966 for topics and 0.699 for subtopics.