Análisis de variables asociadas al rendimiento académico en cursos universitarios virtuales

Abstract

The present study examines strategies and tools that can be used by higher education institutions to identify the most relevant variables associated with academic performance on online courses. A database is created of students who have taken a series of four online university courses. Seven machine learning models are built and their performances are assessed to identify the most optimal. The results show that the random forest model has the best scores in all metrics (accuracy, F1-score, recovery, and precision). Student final grades are associated with age, unsatisfied basic needs index, gender, credits earned, total clicks, region (geographic location), and having or not a disability. In conclusion, it is advisable for universities that offer online courses to create virtual spaces that would let professors learn student background, in addition to the information obtained through regular student interactions with the online educational platform.