Title: Metodología de alerta temprana para el abandono y el rendimiento académico en grados y másters universitarios |
Authors: Sobrino Poveda, Esther |
Tutor: Rabasa, Alejandro  Esteve Campello, Miriam |
Editor: Universidad Miguel Hernández de Elche |
Department: Departamentos de la UMH::Estadística, Matemáticas e Informática |
Issue Date: 2021-09 |
URI: http://hdl.handle.net/11000/26268 |
Abstract:
Son muchos los motivos que conducen a que los estudiantes universitarios acaben abandonando sus carreras. La casuística que conduce al abandono es diferente en función de las ramas académicas, las facultades o escuelas y las circunstancias personales de los universitarios. También es diferente el m... Ver más
There are many reasons that lead college students to end up abandoning their careers. The casuistry that leads to dropout is different depending on the academic branches, the faculties or schools and the personal circumstances of the university students. The timing of this is also different. In a context of continuous changes and an excess of information, Universities consider the possibility of extracting early warning patterns from their databases that help to act in time in those cases where the probability of abandoning begins to be relatively high. This study proposes a two-stage methodology to extract such dropout patterns. In the first place, and depending on each faculty or university school, the model will extract the most influential factors on dropout and later, classification trees will be generated using these explanatory variables. The proposed methodology has been tested with real data from the University Miguel Hernández of Elche, with data from Degrees from the year 2010 to the present (24,894 records), provided directly by the Vice-Rector’s Office for Students affairs and coordination, that leads this project. With the computational experience that accompanies the described methodology, predictive models are capable of modelling dropout with mean accuracies close to 80%. This not only offers a real photograph of the situation of each faculty, but is supported by more than enough confidence so that the Vice-Rector’s Office can design contingency plans for scenarios where dropout is presented as a very likely alternative, in case of do not intervene early
|
Keywords/Subjects: clasificación selección de atributos alerta temprana abandono de estudiantes machine learning classification feature selection early warning students dropping |
Knowledge area: CDU: Ciencias sociales: Demografía. Sociología. Estadística: Sociología. Comunicación |
Type of document: info:eu-repo/semantics/bachelorThesis |
Access rights: info:eu-repo/semantics/openAccess Attribution-NonCommercial-NoDerivatives 4.0 Internacional |
Appears in Collections: TFG - Estadística Empresarial
|