Dropout Prediction in Distance Learning Courses: a Decision Tree-Based Model Integrating Exogenous and Endogenous Causes

Abstract

Although most HEIs are developing programs to control dropout rates, many existing predictive models are based on data or causes that do not directly impact the high dropout rates. In this context, and considering recent studies that recommend the use of data science to develop predictive models, this research aimed to propose a predictive model for dropout in undergraduate EAD courses. Using machine learning and a supervised learning algorithm called decision tree, the model integrated exogenous and endogenous causes of dropout. The research, applied in nature and qualitative in approach, was classified as exploratory and included a case study and literature review. The results revealed the importance and interaction between the variables of engagement, age, and the distance between the student's residence municipality and the EAD center. These findings provide valuable insights for retention strategies in institutions offering distance education.

Keywords: Distance education. Dropout. Predictive models. Data science. Supervised learning.