Machine Learning y predicciones para la mejora del rendimiento en MOOCel caso de la Universitat Politècnica de València
- Martinez Navarro, Jorge Angel
- Linda Johanna Castañeda Quintero Directora
Universidad de defensa: Universidad de Murcia
Fecha de defensa: 05 de mayo de 2021
- Miguel Ferrando Bataller Presidente/a
- María del Mar Sánchez Vera Secretaria
- Francesc Marc Esteve Mon Vocal
Tipo: Tesis
Resumen
The aim of this research is to propose a solution to a specific need raised by a university institution, the Universitat Politécnica de Valencia, which aims to improve the experience of its MOOC (Massive Open Online Course) platform and reduce dropout rates in its courses, through the use of the learning analytics available and machine learning mechanisms. The data analyzed correspond to 700,000 participants spread over 260 courses from 2015 to 2019. To solve this problem, a simple data mining work has not been carried out in which all the possible existing data have been combined, but an investigation based always on pedagogical decisions has been carried out in which, the educational design research methodology is used in order to design automated mechanisms that improve the performance of these courses, through three iterations with different patterns that always end with the presentation of results and feedback of the experts of the university. The main conclusions of this work indicate that, of the 25 pedagogical indicators of drop-out referred to by the bibliographical reviews, only 10 of them are validated with the UPV courses (there is no automatic or automatable data for the others), and of those finally only six of them are possible predictors of student dropout. Finally, a set of automated mechanisms are proposed to be applied in the university's edX platform, in order to improve the user experience and reduce the dropout rate in the courses.