Clustering Gaya Belajar Mahasiswa dengan Metode K-Means: Analisis VARK untuk Pengembangan Strategi Pembelajaran Adaptif

Abstract

This study aims to map students’ learning styles using the VARK (Visual, Auditory, Reading/Writing, Kinesthetic) framework through an unsupervised learning approach. Data were collected from student questionnaires and pre-processed by extracting coded responses, transforming them into numerical variables via one-hot encoding, and applying normalization. The K-Means Clustering algorithm was then employed to group students based on response patterns, with the number of clusters set to four in accordance with the VARK theoretical framework. The results reveal four clusters with distinct characteristics: Visual–Auditory, Auditory–Kinesthetic, Reading/Writing, and Multimodal. Internal validation using the Silhouette Score, Davies–Bouldin Index, Calinski–Harabasz Index, and the Elbow Method confirmed that four clusters represent the optimal configuration. PCA visualization and the distribution of VARK preferences further support the separation among clusters while highlighting the heterogeneity of student learning styles. These findings have practical implications for the design of adaptive learning strategies in higher education. Each cluster requires differentiated approaches, such as the use of visual materials, discussions, hands-on practice, or a variety of methods for multimodal learners. Future studies are recommended to expand the sample size, compare alternative clustering algorithms, and integrate VARK questionnaire data with digital learning behavior to enrich the analysis.