Estudo da evasão de alunos de graduação utilizando Educational Data Mining

Abstract

The problem of evasion is the object of study of several areas and a recurring concern in Federal Institutes of Higher Education in Brazil, because it is associated with the social and resources losses of all those involved in the teaching process. The agile and conclusive analysis of data that the Institutions have is extremely important to carry out preventive actions for the problem. In this sense, this work aims at the study of dropout using the EDM of the Computer Science program of the Federal University of Pampa (Unipampa), under two aspects: the analysis of the socioeconomic profile of the entrants through the data SiSU/ENEM, referring to the years 2010 to 2018 and the situation of the students enrolled in the first year in the curricular components referring to the first two semesters in the time axis from 2009 to 2018.The analyses of the entrants and enrolled students profiles were done separately and therefore it was not possible cross the information. In the socioeconomic analysis it was found that even if the dropout in general has a high index, there are groups of higher risk that need a closer look of the Institution. The mark in the competences by area of evaluation ENEM/SiSU, does not have such a considerable impact on the permanence of this student. In the analysis of the profile of the student of the program, it was demonstrated that there is a pattern that can be mapped using the predictive models explained in this work, with the use of supervised methods obtaining an excellent accuracy.