Personalized E-Learning Recommender System Using Machine Learning

Various domains, including e-commerce, entertainment, and education, employ recommender systems (RSs) to offer personalized suggestions based on user preferences, needs, and goals. In education, RSs assist learners in discovering suitable resources, activities, and paths aligned with their learning styles, knowledge levels, and objectives. However, designing and evaluating effective educational RSs is challenging, requiring the consideration of several factors like the learner’s profile, learning context, pedagogical principles, and learning outcomes. The integration of machine learning with multi-agent systems enhances the effectiveness of these systems. Our research introduces a multi-agent system (MAS) that classifies students to recommend courses based on their knowledge levels and learning styles. Comprising three agents—the learner agent, the tutor agent, and the course agent—the system employs the Felder and Silverman model to identify different learning styles. Support Vector Machine (SVM) and K-nearest neighbors (KNN) algorithms were implemented and compared using metrics such as accuracy, precision, recall, and F1-score. The study found that KNN outperformed SVM, demonstrating superior performance in classifying learners based on preferences, learning styles, and knowledge levels. This classification capability enables the system to recommend courses tailored to individual needs, promoting increased engagement and efficiency.