Predicting student learning performance to suggest courses is a vital role of an academic adviser in the Intelligent Tutoring System (ITS) as well as the university's E-learning system.Many different approaches, such as classification, regression, association rules, and recommender systems, have been used to solve this problem. Recently, using collaborative filtering in the recommender system, particularly the matrix factorization technique, to develop the courses' recommendation system was a measurable success. Many breakthroughs have been made to increase prediction accuracy, such as leveraging student profiles, course features, or course relationships, but they have not yet been mined. This paper suggests a method for improving prediction accuracy by including course relationships into the course recommendation system. When we validate the published educational datasets, the experimental outcomes of the proposed approach are positive.