Abstract
Transitioning from pre-university studies to a bachelor's degree can be quite challenging, as students often have to choose from a wide range of programs without knowing their academic compatibility. This lack of information can lead to poor performance or even dropout. To tackle this issue, we conducted a study at a Spanish university using machine learning (ML) algorithms on academic data from 2010 to 2022 (about 72,000 records) to develop a degree recommendation tool aligned with pre-university profiles. Our results show an average accuracy of 70% for the top 5 predictions and 90% for the top 10. Moreover, explainability techniques allowed us to identify profiles according to bachelor's degree programs and observe relationships between pre-university subjects and chosen degrees. For example, students who take Geography in their access proofs are less likely to choose Computer Engineering, while Mathematics, English, and Physics negatively affect recommendations for Education degrees. The tool is designed to assist school counselors by providing comprehensive and accurate guidance, considering students' academic profiles, interests, and socioeconomic factors. This is expected to improve academic performance and reduce dropout rates. Future work includes expanding the number of academic records, incorporating additional universities, and introducing new ML algorithms to enhance our results.