Aplicación de Técnicas de Inteligencia Artificial para la Predicción del Rendimiento Académico en Asignaturas de Matemáticas.

Saved in:

Bibliographic Details
Title:	Aplicación de Técnicas de Inteligencia Artificial para la Predicción del Rendimiento Académico en Asignaturas de Matemáticas.
Alternate Title:	Application of Artificial Intelligence Techniques for Predicting Academic Performance in Mathematics Courses.
Authors:	LONDOÑO, MAVELYN STERLING¹, BARBOSA GÓMEZ, FABIO ANDRÉS¹, BEDOYA LEIVA, OSCAR FERNANDO¹ oscar.bedoya@correounivalle.edu.co
Source:	Revista EIA. ene-jun2026, Vol. 23 Issue 45, p1-25. 25p.
Abstract (English):	Low academic performance in mathematics courses in higher education is a recurring cause of failure and student dropout, particularly in science, technology, and engineering programs. This article presents the development of a predictive model based on machine learning techniques, aimed at early identification of students at risk of failing such courses. An institutional dataset comprising information from 2,932 students admitted to Universidad del Valle between 2021 and 2023 was used. The dataset includes demographic, socioeconomic, and academic variables, as well as the results of a mathematics diagnostic test. Neural networks, decision trees, and support vector machines were trained and evaluated using preprocessing, class balancing, and cross-validation techniques. The neural network model achieved the best performance, with an AUC of 70.6%, accuracy of 64.9%, sensitivity of 64.4%, and specificity of 65.1%, outperforming decision tree models (AUC of 46.7%) and SVMs (AUC of 66.1%). Compared to previous studies, the results are competitive given the heterogeneity of the data and the model's practical integration. Unlike other approaches that rely on data generated during the semester, this model operates solely with information available at the time of admission, enabling timely interventions before the academic term begins. The solution was integrated into the institutional ASES system through an API, allowing real-time prediction queries and the generation of personalized alerts. As such, it serves as a valuable tool for strengthening academic support strategies and enhancing student retention in high-risk areas such as mathematics. [ABSTRACT FROM AUTHOR]
Abstract (Spanish):	El bajo rendimiento en asignaturas de matemáticas en la educación superior es una causa recurrente de reprobación y deserción estudiantil, especialmente en programas de ciencias, tecnología e ingeniería. Este artículo presenta el desarrollo de un modelo predictivo basado en técnicas de aprendizaje automático que busca identificar tempranamente a los estudiantes en riesgo de no aprobar estas asignaturas. Para esto, se utilizó una base de datos institucional con información de 2932 estudiantes admitidos en la Universidad del Valle entre 2021 y 2023, que incluye variables demográficas, socioeconómicas, y el resultado de una prueba diagnóstica de matemáticas. Además, se entrenaron y evaluaron modelos de redes neuronales, árboles de decisión y máquinas de soporte vectorial, aplicando técnicas de preprocesamiento, balanceo de clases y validación cruzada. El modelo de red neuronal mostró el mejor desempeño con un área bajo la curva ROC (AUC) de 70.6%, exactitud del 64.9%, sensibilidad del 64.4% y especificidad del 65.1%, superando a los modelos de árboles de decisión (AUC de 46.7%) y SVM (AUC de 66.1%). En comparación con estudios previos, los resultados son competitivos considerando la heterogeneidad de los datos utilizados y el enfoque práctico del sistema. A diferencia de otras investigaciones que requieren registros generados durante el semestre, este modelo utiliza únicamente información disponible antes del inicio de clases, lo que permite intervenir desde etapas tempranas. La solución se integró al sistema institucional ASES a través de una API, lo que posibilita la consulta de predicciones en tiempo real y la emisión de alertas personalizadas. De este modo, se constituye en un recurso valioso para fortalecer las estrategias de acompañamiento académico y fomentar la retención estudiantil en asignaturas de alta dificultad como las matemáticas. [ABSTRACT FROM AUTHOR]
	Copyright of Revista EIA is the property of Escuela de Ingenieria de Antioquia 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.)
Database:	Engineering Source

Description
Abstract:	Low academic performance in mathematics courses in higher education is a recurring cause of failure and student dropout, particularly in science, technology, and engineering programs. This article presents the development of a predictive model based on machine learning techniques, aimed at early identification of students at risk of failing such courses. An institutional dataset comprising information from 2,932 students admitted to Universidad del Valle between 2021 and 2023 was used. The dataset includes demographic, socioeconomic, and academic variables, as well as the results of a mathematics diagnostic test. Neural networks, decision trees, and support vector machines were trained and evaluated using preprocessing, class balancing, and cross-validation techniques. The neural network model achieved the best performance, with an AUC of 70.6%, accuracy of 64.9%, sensitivity of 64.4%, and specificity of 65.1%, outperforming decision tree models (AUC of 46.7%) and SVMs (AUC of 66.1%). Compared to previous studies, the results are competitive given the heterogeneity of the data and the model's practical integration. Unlike other approaches that rely on data generated during the semester, this model operates solely with information available at the time of admission, enabling timely interventions before the academic term begins. The solution was integrated into the institutional ASES system through an API, allowing real-time prediction queries and the generation of personalized alerts. As such, it serves as a valuable tool for strengthening academic support strategies and enhancing student retention in high-risk areas such as mathematics. [ABSTRACT FROM AUTHOR]
ISSN:	17941237
DOI:	10.24050/reia.v23i45.1933