Pengembangan Model Rekomendasi Pembelajaran Berbasis Gaya Belajar Menggunakan Pendekatan Multiple Intelligences Theory
Gerzon Jokomen Maulany, Prof. Ir. Paulus Insap Santosa, M.Sc., Ph.D., IPU.; Dr. Indriana Hidayah, S.T., M.T.
2026 | Disertasi | S3 Teknik Elektro
The rapid development of e-learning over the past two decades has fostered the emergence of adaptive learning approaches to enhance the relevance and effectiveness of instructional processes. However, most learning-style–based adaptive learning models still assume a single dominant learning style, even though learners may simultaneously exhibit multiple learning tendencies. Multiple Intelligences theory provides a more representative framework for capturing this diversity, making its integration into adaptive learning systems a critical requirement. In response to limitations identified in prior studies—such as the use of multi-class approaches incompatible with multi-label characteristics, limited and imbalanced datasets, and the lack of systematic integration with recommendation systems—this research is designed to develop and validate a data-driven, adaptive learning recommendation model grounded in Multiple Intelligences theory.
This study employs a semi-supervised learning approach to detect learners’ multiple intelligences profiles in a multi-label manner based on students’ interaction patterns within the Learning Management System (LMS). The model is developed and validated using datasets from two different institutions, with the first dataset used for model training and the second for testing in order to assess generalization capability. The resulting multiple intelligences profiles are subsequently utilized as the basis for developing an adaptive learning recommendation system through content-based filtering (CBF), collaborative filtering (CF), and hybrid filtering (HF) approaches.
The results demonstrate that LMS interaction patterns—comprising activity types, access frequency, and learning duration—play a significant role in differentiating MI tendencies. The proposed semi-supervised approach achieves an accuracy of 89% and successfully identifies seven learning styles simultaneously, outperforming prior studies that generally detected only five. Evaluation of the recommendation system indicates that hybrid filtering yields the best performance, with a Precision@5 of 0.36, a Recall@5 of 0.90, a Mean Reciprocal Rank (MRR) of 0.84, and a Normalized Discounted Cumulative Gain (NDCG) of 0.85. A compliance rate of 36.7% suggests that the effectiveness of recommendations is influenced by the suitability and presentation of the learning materials, as students who followed the recommendations demonstrated higher learning frequency and duration. Explicitly, this study contributes by developing a semi-supervised multi-label approach for detecting Multiple Intelligences learning styles that is directly integrated into an LMS-based learning recommendation system, thereby enabling more adaptive, inclusive, and context-aware learning personalization in higher education.
Kata Kunci : Sistem Rekomendasi Pendidikan, e-learning, Model Pembelajaran, Gaya Belajar, Kecerdasan Majemuk, Klasifikasi Multi-Label, Semi-supervised, Hybrid Filtering.