The project “Tecnologías de modelado dinámico de estudiantes y asistentes digitales para la mejora de resultados en plataformas de e-learning” (Technologies for the dynamic modeling of learners and digital assistants for performance enhancement in e-learning platforms) WITH_YOU (2022-2025), managed by Editores y Consultores Especializados (ECE), with the participation of the Universidad Europea del Atlántico (European University of the Atlantic, UNEATLANTICO), and the Polytechnic University of Catalonia (UPC) is currently in its second year of execution.
During the first year, the project work focused, on the one hand, on the definition of a data capture system, for which an analysis was made of the student variables (socio-demographics, academic results, interactions, etc.) that, a priori, were considered determinant in the academic progress of the students enrolled in the e-learning platforms and, on the other hand, on the development of a data capture model that would allow the extraction of all the required information from the platforms for their future exploitation.
Additionally, the first steps have also been taken to define a satisfaction survey module to enrich the information extracted (mining) from the platform. As a result of this work, it has been possible to obtain a database with more than 280,000 records and more than 160 features that describe the academic performance of students pursuing degrees in e-learning platforms.
The second year’s work has taken as a starting point the results of the previous year, for which a preprocessing of the ensemble data to feature engineering has been carried out on the initial database. This work has made it possible to select and/or generate 40 preprocessed variables, which contain the behavioral information (in this case, the academic progress) of more than 34,000 unique users of the platforms during the time in which the data were being mined.
Currently, work has begun to exploit the latent information within these variables. Clustering technologies have begun to be developed using neural network and distance-based algorithms to create a dynamic student modeling system that considers the temporal progression of students in their degrees. Preliminary results from these algorithms reflect encouraging data, as performance metrics show that the systems correctly group eight out of ten students based on the training parameters set for the models. These results suggest to the research team that, after a process of parameter optimization or selection of new algorithms, it may be possible to improve the initial results, which is very favorable for decision-making.
On the part of the UPC, those responsible are researching and developing technologies for social profile analysis. Considering the relevance of the social profile (including interactions between students and teachers) and the availability of different graph analysis techniques that should allow its modeling, this specific task is devoted to the automated analysis of the students’ social profile.
For the remainder of the year and for the last year of the project, work will focus on optimization, development of an intervention system, and prototyping and validation of the proposed solutions.
The initiative is framed within the state research plan 2021-2023 within the call for Public-Private Collaboration Projects 2021.
The publication is part of project CPP2021-008349, funded by MCIN/AEI/10.13039/501100011033 and by the European Union-NextGenerationEU/PRTR”.