Download PDFOpen PDF in browser Switch back to the title and the abstract in Spanish Prediction of University Students at Academic Risk Using Supervised AlgorithmsEasyChair Preprint no. 661010 pages•Date: September 16, 2021AbstractThe purpose of this work was to create predictive models using Supervised Classification Algorithms, in order to make known that students were at academic risk and to be able to carry out a focused follow-up. In this study, the CRISP-DM methodology was used to create predictive models, taking full advantage of the data obtained by the university itself, where these only contain academic qualifications. Some important findings obtained during the data analysis was the importance of the summer period, thanks to this cycle the number of students at risk decreases significantly. Furthermore, the majority of at-risk students are focused on the first four semesters. Five classifiers are presented, Bayesian Classifier, Artificial Neural Network, Discriminant Quadratic Analysis, Support Vector Machine and Logistic Regression. The choice of the best model is based on two Performance Measures, the ROC Curve and Sensitivity, then the two best models are presented according to the resources that the institution has, the Bayesian Classifier when there are enough resources and the Logistic Regression when resources are scarce. Keyphrases: Algoritmos Supervisados, Calificaciones Académicas, Clasificador Bayesiano, Sensibilidad
|