Profiling student performance for multi-agent personalization in virtual reality

Abstract

This study uses the open university learning analytics dataset (OULAD) to cluster student performance data to improve personalized learning. Three main aspects are the focus of the analysis: instructional involvement, behavior, and demographics. To create significant, comprehensible student profiles, the clustering algorithms k-means, k-modes, and k-prototypes were used for each dimension independently. In order to forecast student categories from input features, supervised classification models, such as support vector machines (SVMs) and random forests, were trained using these profiles as targets. Accuracy, F1-score, and cross-validation were used to assess the categorization models' performance. The outcomes demonstrate how well unsupervised and supervised learning strategies may be combined for adaptive learning. These profiles serve as a foundation for the future design of a multi-agent virtual reality (VR)-learning environment. In this envisioned system, specialized agents would handle behavioral adaptation, demographic personalization, and pedagogical coordination, offering a personalized learning experience tailored to each learner’s profile.