Abstract
School dropout has been a long-standing issue in Higher Education
Institutions (HEIs), raising concerns and prompting mitigation
efforts. This paper examines dropout in a Technology course at
a Federal Institute, addressing the research question: "How can
machine learning classifiers assist in predicting student dropout?".
The study applies machine learning classifiers using data from a
Federal Institute, including demographic (age group, income, city,
special needs, ethnicity, gender) and academic variables (admission
method, vacancy type, enrollment status). The objective is to
develop analyses ranging from descriptive statistics to predictive
modeling. The classifiers used include Support Vector Machine, Random
Forest, Logistic Regression, Gradient Boosting, AdaBoost, and
K-Nearest Neighbors. The goal is to support teachers, coordinators,
and administrators in implementing preventive measures such as
personalized mentoring, continuous monitoring, and institutional
policies to improve academic infrastructure and inclusivity. The
results show that Support Vector Machine, Gradient Boosting, and
AdaBoost achieved the best performance, with F1-measure values
between 0.6 and 0.8, demonstrating their predictive capability. This
highlights the potential of machine learning in addressing student
dropout in higher education.