Treffer: Predictive model for the classification of university students at risk of academic loss.

For higher education institutions, predicting the risk of academic loss is a priority issue due to the resources invested by institutions, students and the academic community in general. Objective: the objective of this research was to propose a suitable model that allows predicting students who are at risk of academic loss in a chemistry course. Methodology: the quasi-experimental, predictive, longitudinal research was developed with data from 103 students from four Colombian universities. To build the model, a comparison of five algorithms was implemented. Data was processed with Jupyter-Python. Results: the logistic regression model (LR) was built based on the students' results on the Saber 11 test (Colombian nation-wide university admission exam), in which the penalty of false positives with different weights from the false negatives improved the performance of the model. Conclusions: it is concluded that LR is substantially better than grasping or a guessing approach, furthermore, it was shown to perform better than a neural network model. [ABSTRACT FROM AUTHOR]

Para las instituciones de educación superior, predecir el riesgo de pérdida académica es un tema prioritario debido a los recursos invertidos por las instituciones, los estudiantes y la comunidad académica en general. Objetivo: el objetivo de esta investigación fue proponer un modelo adecuado que permita predecir a los estudiantes que están en riesgo de pérdida académica en un curso de química. Metodología: la investigación cuasi-experimental, predictiva y longitudinal se desarrolló con los datos de 103 estudiantes de cuatro universidades colombianas. Para construir el modelo se implementó una comparación de cinco algoritmos. Los datos se procesaron con Jupyter-Python. Resultados: el modelo de regresión logística (LR) se construyó con base en los resultados de los estudiantes en la prueba Saber 11 (examen nacional colombiano de admisión a la universidad), en el que la penalización de falsos positivos con pesos diferentes a los falsos negativos mejoró el rendimiento del modelo. Conclusiones: se concluye que LR es sustancialmente mejor que un enfoque codicioso o de adivinanzas, además, se demostró que funciona mejor que un modelo de red neuronal. [ABSTRACT FROM AUTHOR]

Copyright of Educación y Humanismo is the property of Universidad Simon Bolivar and its content may not be copied or emailed to multiple sites without the copyright holder's express written permission. Additionally, content may not be used with any artificial intelligence tools or machine learning technologies. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)