MACHINE LEARNING TO PREDICT STUDENT ACADEMIC RISK IN ENGINEERING

Resumen

Identifying academic risk in an engineering program is very important since it allows to stablish policies that can be applied to prevent it and to guarantee student success. At Pontificia Universidad Javeriana in Bogotá – Colombia, we have developed some methods that can be used to predict the number of students that are in academic risk or even in dropout risk. This information can be employed to identify possible actions and interventions in our programs. We focus in the necessity to design and implement models to recognize patterns associated with academic performance and transitions of undergraduate students.

Our strategies are faced to accompany the students. Once we identify the risk, every student is evaluated by an academic adviser, who is in charge of determining the possible causes of the low performance and this way customize the support. In this sense, weaknesses are identified around CDIO competencies, for example, interpersonal competences, basic knowledge, algorithmic thinking, problem solving or personal skills. In this way, students are routed to different strategies to strengthen those abilities.

This paper starts with an introduction justifying this work. After that, a section describes the methodology to predict academic risk. Then, we present our accompaniment strategies. Some future works are also presented. In this context, this work tries to give hints about particular intervention for student in risk.