Hybrid Machine Learning for Personalized Course Recommendation and Conflict Free Academic Schedule Generation

Course registration at many universities is harder than it needs to be. Students face overlapping prerequisites and scheduling conflicts with limited advising, which leads to poorly sequenced courses and avoidable failures. EduTrack, developed and evaluated in the University of Jordan’s Computer Engineering program, combines XGBoost multi-label classification with content-based filtering to recommend courses from each student’s academic history. A constraint-aware algorithm turns these recommendations into a conflict-free semester schedule. The web application runs on Python and Flask, with Google Cloud Firestore for storage. The classifier reached a micro-averaged F1-score of 77%, reasonable for the small dataset, but augmentation created near-duplicate records with correlated labels, so the figure is probably optimistic. A concurrent think-aloud usability study with 235 participants assessed comparison with the current process, trust in the recommendations, schedule quality, and likelihood of use. Responses were positive on all four, and students found the system easier to navigate.