El número de cursos on-line ha crecido sustancialmente en Mondragon Unibertsitatea en los últimos años http://www.mondragon.edu/cursos/es/cursos-online y es por ello que la extracción automatizada de información relativa al uso que el estudiante a distancia hace de este curso es una pista valiosa tanto para el propio alumno como para el profesor del curso.
A continuación tenéis la oportunidad de saber sobre un proyecto cuyo objetivo es aprender automáticamente sobre como los buenos alumnos aprenden del contenido disponible en un curso para posteriormente facilitárselo al nuevo usuario como sugerencia o posible pauta a la hora de consumir los recursos disponibles. Los resultados esperados serán mejorar los resultados aumentando a la vez su compromiso.
Recommender systems are an important part of the information for e-commerce ecosystems and also for e-learning platforms. They help the user filtering through large information and product spaces and have been widely studied in the last decades. A project named MY_COURSE: Personalización en la presentación de contenidos educativos was carried out for the last year and a half, funded by Rural Development and Tourism Innovation Department of Gipuzkoa Provincial Council. The research uses this technology embedded in learning domains paying attention to the specific student needs. Using association rules based on resources consumed by the best students of the previous editions of this course, a dynamic changing environment suggests the most valuable resources that can help new users better learn how to progress in a given course. Students will learn easier using this information, that will help them to discriminate resources to consume.
Advanced data mining tasks have be used to dynamically provide the platform with this information. Specifically, the platform uses association rule mining in educational resources in order to find the likelihood of resources co-occurrence. This information is tailored specifically for each student and is provided to them every week by updating the recommendations in a new block designed for this purpose, which is highlighted in the next image.
The core for recommender systems are the users and their preferences expressed as a triplet, like (user, item, rating). For us, the rating is obtained by counting the clicks done to the resource by the students with the best scores. After the rules are generated for a given delivery of the course, when a new one starts, these rules are used as predictions for the new students.
It is true that the courses will evolve and in this sense, the rules will also evolve. To cover these problems, rules will be labelled as “new”, “alive” or “successful” or “deprecated”. This way the dynamic nature of the course design and technology used will be absorbed. Next steps will be to widely evaluate the recommender system.
These type of systems want to fill the gap for the new students using the platform as if they were always new and provide them valuable information to learn faster and learn better, learning from previous experiences.