Abstract <p>The expansion of e-learning has transformed modern education, creating opportunities to analyze student behavior and improve learning outcomes. Since students exhibit unique learning styles shaped by their behaviors, accurately identifying these styles is essential for delivering personalized resources. However, the vast amount of behavioral data presents challenges in effectively capturing student interactions. Moreover, many existing models focus solely on individual learner attributes, often overlooking the interaction between students and educational resources. To address these limitations, this study introduces a Density Peak Clustering (DPC) and Support Vector Machine (SVM) framework for e-learning style prediction. DPC is employed to group learners based on behavioral similarities, forming meaningful clusters of learning styles. SVM is then used to classify these clusters, enhancing prediction accuracy and improving e-learning efficiency. Experimental validation using the Open University Learning Analytics Dataset (OULAD) confirms the effectiveness of this approach in identifying diverse learning styles, supporting the development of more adaptive and personalized e-learning environments.</p> Graphical Abstract <p></p>