RESUMO
School dropout is a matter of great relevance due to its significant
social, economic, and personal repercussions. This study focuses
on the issue of course attrition in higher education, a significant
contributing factor to dropout rates. Here, we present the analysis
of 216 results, obtained from combining nine different preprocessing
and three sampling methods, as well as eight machine learning
algorithms, based on student attendance, grades, and completion of
evaluative activities, extracted from the MOODLE learning environment.
The results of this study, obtained through implicit analysis,
reveal a combination of methods capable of identifying student
attrition with a recall of 95.31% and accuracy of 92.67% in just 25%
of semester’s duration.