Investigando indicadores de evasão em cursos superiores EAD utilizando análise de dados e aprendizado de máquina

Abstract

School dropout in the context of Distance Learning has become an increasing concern in Higher Education. In this study, using exploratory analysis, dropout patterns were identified in 36 distance learning undergraduate courses offered by a non-profit University, considering geographic, personal and academic aspects. The most relevant variables for classifying classes as dropouts were also determined, using the decision tree, random forest and XGBoost algorithms. The results, in addition to providing patterns for the aspects investigated, also revealed that time-related characteristics have greater transparency with resistance. Furthermore, the provisions state that variables such as the average final grade, the number of failures, the type of secondary education institution attended by the student and the location of the teaching center play a crucial role in determining a possible student dropout, according to the classified decision tree and XGBoost, which obtained the best performance results. These analyzes provided a clearer view of the factors that influence dropout in Distance Learning, allowing preventive and personalized measures to be applied to improve retention students and enhance the academic experience.