Attention-Based Neural Networks with TF-IDF Features and BERT Embeddings for Enhanced Course Recommendation System

One vital component of the curricula-variable system is course recommendation aiming to identify appropriate courses for students depending on their needs. But when it comes to choosing the best courses, students regularly encounter a cognitive overburden. Predicting the relevance of courses is essential for educational systems. This research enhances text representation for such predictions by integrating TF-IDF features and BERT embeddings. By merging TF-IDF features and BERT embeddings, this research proposes a multi-Head attention mechanism integrated into a feedforward network as a solution to this problem. The effectiveness of the proposed recommendation model is evaluated and compared against a Convolutional Neural Network (CNN) and a Long Short-Term Memory (LSTM) model. It has been found that the proposed model consistently outperforms in terms of ranking quality over a range of recommendation list sizes. NDCG@k, ARHR@k, F1-score, accuracy, precision, recall, and other performance indicators are used to build and evaluate each learning model.