Estimación de Riesgo de Fallo en Materias Basado en Similitud y Manejo de Incertidumbre Dada por Datos Académicos Históricos

Abstract

The academic drop-out is not a problem that affect just to students, this affects universities as well, because failing to complete an educational degree is an inefficient use of resources for both students and the university; one of the factors related with the drop-out is the interpersonal competence, which allows a student establish a relationship with him/her teachers and classmates, and this could affect/contribute to the successful culmination of him/her academic career; ESPOL’s Academic Counselling service, allows students to establish a relationship with their academic adviser, who provides guidance to improve the academic performance; the help provided by the advisers are based on their expertise and the student’s academic information, but this process can be improved by adding “educated” information; for example, an estimation factor of risk for flunking a course in a semester. The following paper proposes an architecture for unsupervised classification based on clustering and fuzzy prototyping, then assessing the risk via extraction of descriptive variables. A prototype was developed implementing the proposed model, which have three principal components: pre-processer, clustering and classifier; this software allowed to validate the model in order to predict the risk, based on the academic load and performance, with a significant certainty performance.