Statistical and connectionist models for predict the academic performance of universitary students

This paper analyzes the relationship between the academic performance of students entering professional profile's careers in the FACENA - UNNE in Corrientes, Argentina, during the first year, and their social-educational characteristics.Performance was measured by the approval of the partial ev...

Descripción completa

Guardado en:
Detalles Bibliográficos
Autores principales: López, María V., Longoni, María G., Porcel, Eduardo A.
Formato: Artículo revista
Lenguaje:Español
Publicado: Escuela de Perfeccionamiento en Investigación Operativa 2018
Materias:
Acceso en línea:https://revistas.unc.edu.ar/index.php/epio/article/view/20348
Aporte de:
Descripción
Sumario:This paper analyzes the relationship between the academic performance of students entering professional profile's careers in the FACENA - UNNE in Corrientes, Argentina, during the first year, and their social-educational characteristics.Performance was measured by the approval of the partial evaluation of the subjects in the first semester of the first year. A model of Multinomial Logistic Regression (MLR) and two models of neural networks of type Multilayer Perceptron (MP) and Radial Basis Function (RBF) were fitted to two data sets: a) students entering in Biochemistry, whose curriculum includes two subjects in the first semester of the first year, b) students entering careers whose curriculum includes three subjects in the first semester of the first year.In both cases, the PM model produced the best fit, and besides it was observed that in the case b) the three techniques showed high percentages of correct classification. The obtained results contribute to guide policies and strategies to improve the worrying levels of dropout and low performance of students in the first year of college.