Abstract:

Concern for the academic success of an increasingly diverse student body is receiving greater national and global attention in higher education. Given the increasing exigencies and evolving challenges that students now face, higher education institutions are called upon to develop multifaceted solutions that allow for the identification and prevention of academic pathways that may put their students at risk of failing or dropping out. This paper presents the results of a trial carried out at a Portuguese public university involving the development of digital tools, using machine learning models to help bolster efforts aimed at mitigating the risk of dropout and failure in higher education. From the outset, this trial has been built on interdisciplinary cooperation between specialists within the social sciences, information systems and information technology as well as between teachers, students, and various university departments (Educational Management, Social Action, Soft Skills Lab, Pedagogical Council, Computer Science, Information Systems and Quality Management). The creation of these tools is mainly based on the definition and implementation of an internal information system (Fenix) currently in the testing phase.

Implementation has followed a multi-stage process guided by the following objectives and procedures:
a) defining success and dropout indicators by national and international guidelines;
b) identifying the scale of the problem within the institution;
c) understanding the patterns of the problem under study;
d) identifying critical factors of failure and dropout;
e) to implement a digital alarm system that contributes to preventive action and improves conditions for success, especially focused on students who are at risk of failure or dropout.

It is expected that such a system will automatically identify pathways to both success and failure, enabling accurate and comprehensive analysis of students' academic data, including patterns and key indicators that can be used to predict risks of failure, while enabling proactive and personalized interventions to improve performance and maximize opportunities for academic success.

Keywords:

Machine learning models, alarm systems, dropout, higher education students.