Dropout or permanence? Predictive models for higher education management

Abstract

This research aimed to propose statistical predictive models for the dropout management in undergraduate courses of a Brazilian higher education institution. For this, we conducted an applied study in four undergraduate e-learning courses at a Brazilian public university. We collected the data of 2,991 students from the university’s institutional systems and we used the binary logistic regression method. In the end, we conclude that for different courses, different variables can influence the dropout phenomenon, as well as the same variable can generate different effects in different realities. In addition, the statistical predictive models developed allowed the inference “dropout” or “permanence” for active students at the time of data collection. In partial assessment of the accuracy of the models, we identified that 9 out of 10 dropouts that occurred were previously identified by the models developed. This highlights the potential for using predictive models to the student dropout management, providing a basis for reviewing educational policies and management by identifying the variables that influence student dropout and permanence, as well as by early identification of students at risk of dropout.