Pengembangan Sistem Adaptive Learning Berbasis Bayesian Knowledge Tracing dan Ontologi untuk Matematika SD

Abstract

Pembelajaran matematika menghadapi tantangan akibat heterogenitas kemampuan siswa dan keterbatasan pendekatan pembelajaran konvensional dalam mengakomodasi kebutuhan individual siswa. Adaptive learning menjadi salah satu solusi karena mampu menyesuaikan materi berdasarkan tingkat penguasaan siswa. Namun, sebagian besar sistem adaptive learning masih memiliki keterbatasan dalam merepresentasikan hubungan antarkonsep pengetahuan secara eksplisit. Penelitian ini bertujuan mengembangkan sistem adaptive learning berbasis Bayesian Knowledge Tracing (BKT) dan ontologi pedagogis untuk pembelajaran matematika Kelas 1 SD. Hasil evaluasi menunjukkan bahwa model BKT + Ontologi memperoleh nilai AUC-ROC sebesar 0,8049, accuracy sebesar 82,06%, dan RMSE sebesar 0,3872. Hasil tersebut menunjukkan bahwa integrasi ontologi pedagogis dan BKT berpotensi meningkatkan kemampuan sistem dalam memodelkan penguasaan siswa dan menghasilkan jalur pembelajaran adaptif yang lebih terstruktur.

Mathematics learning faces challenges due to the heterogeneity of students' abilities and the limitations of conventional learning approaches in accommodating individual learning needs. Adaptive learning offers a promising solution by personalizing learning materials according to students' mastery levels. However, most existing adaptive learning systems still have limited capability to explicitly represent relationships among knowledge concepts. This study aims to develop an adaptive learning system based on Bayesian Knowledge Tracing (BKT) and a pedagogical ontology for first-grade elementary mathematics. Evaluation results show that the proposed BKT + Ontology model achieved an AUC-ROC of 0.8049, an accuracy of 82.06%, and an RMSE of 0.3872. These findings indicate that integrating pedagogical ontology with BKT has the potential to improve student knowledge modeling and generate more structured adaptive learning pathways.