A Hybrid Course Recommender System using LLM-derived embeddings and Neural Collaborative Filtering

Abstract

Personalized education depends on recommender systems assisting students in picking courses according to the preferences and goals they need to learn. In this research, we present a novel hybrid course recommender system that combines the insight from learner and course classification models using collaborative filtering techniques and generative AI.

According to their academic performance and behavioural data, students are categorized into adaptive and accelerated learners using machine learning algorithms. At the same time, Natural Language Processing (NLP) applied to course descriptions, reviews and metadata categorize courses into foundational and advanced courses. Such classifications serve as a basis for a personalized recommendation framework helping learners select proper courses.

Large language models (LLMs) are used to generate semantic relationships between courses via generative AI. Combining these embeddings with collaborative filtering based on student course interaction data, we offer highly accurate and personalized recommendations. The proposed system solves cold start well and makes substantial recommendation precision and user satisfaction gains over baseline methods.

The approach shows how traditional recommendation paradigms can be combined with Generative AI to give a scalable and interpretable suggestion for educational personalization. Future extensions of this system may involve reinforcement learning for real-time adaptation to user preferences.