Predicting Academic Performance in a Subject Using Classifier Algorithms

Peralta E.A.S.; Málaga G.R.C.; Pari S.B.C. · CEUR Workshop Proceedings (2023)

Resumen

The objective of this research is to determine the academic performance of students starting a Systems Engineering course. The discrete structure I course which is considered a course that is difficult to pass. The population is represented by 827 students, the research was approached from a quantitative approach, non-experimental design and at a correlational level. The methodology implemented is CRISP-DM (Cross Industry Standard Process for Data Mining) through the supervised learning technique using binary classification models based on random forest algorithms, xgboost and support vector machines. The results have allowed predicting if a student will pass or fail the course. The classification models that have shown the best results are based on random forest and xgboots algorithms with an accuracy of 82.5%.

Instituciones: Universidad Nacional de San Agustín de Arequipa