Selecting the right learning content according to learners’ learning abilities and interests is the first and most important factor in achieving good learning performance. Based on the similarity between the course rating data in the Collaborative Filtering format (user, item, rating), and along with the development of Graph Neural Networks (GNN) in developing recommendation systems, we tried to develop a Collaborative Filtering (CF) model based on GNN architecture to recommend suitable courses to the learners. In this study, two CF models based on GNN, including Neural Graph Collaborative Filtering (NGCF) and Light Graph Convolutional Neural Networks (LightGCN), were experimentally compared with some traditional CF models such as Regularized Matrix Factorization (RMF), Light Matrix Factorization (LMF), and Neural Collaborative Filtering (NCF). The experimental results on the Coursera Course Review dataset using LightGCN give promising results in the precision, recall, and RMSE metrics.