TRACE: Transformer Model Predicts Courses and Grades Jointly
A recent study presents TRACE (TRansformer for Academic Course-grade Estimation), a model designed to simultaneously forecast both the courses a student will enroll in and their expected grades for the next semester. This method overcomes a shortcoming of current learning analytics models that view academic history merely as a linear sequence, neglecting the overlapping nature of courses within a semester. By organizing courses by semester, TRACE effectively accounts for course concurrency. The model employs an innovative loss function that merges course-set and grade predictions. Utilizing ten years of institutional data, this joint prediction model shows marked enhancements in prediction accuracy compared to those predicting grades in isolation. The research can be found on arXiv under identifier 2608.13409.
Key facts
- TRACE is a transformer-based model for academic course-grade estimation.
- It jointly predicts courses and grades for an upcoming semester.
- Existing models treat academic history as a simple sequence, overlooking course concurrency.
- TRACE encodes courses per semester to capture concurrency effects.
- A novel loss function combines course-set prediction with grade prediction.
- Trained on ten years of institutional data.
- Joint prediction leads to significant improvements in prediction quality.
- Paper available on arXiv:2608.13409.
Entities
Institutions
- arXiv