Aplicação de Modelos de Aprendizado de Máquina na Predição da Desistência Acadêmica em Cursos de TI do IFPE

Abstract

Dropout rates are one of the main challenges faced by educational institutions, especially in the Exact Sciences, which have high dropout rates. This study, applied to the IFPE context, proposes the use of machine learning techniques to predict student dropout rates based on academic, demographic, and socioeconomic data from students between the 2001–2025 academic years. After data preprocessing, which included variable coding, missing value processing, and class balancing with SMOTE, three supervised models were trained: Decision Tree, Extra Trees, and Multilayer Perceptron (MLP). The models were adjusted using Grid Search and evaluated using metrics such as accuracy, precision, recall, and F1-score. The MLP performed best, standing out for its ability to more accurately identify students prone to dropout. The results demonstrate the potential of artificial intelligence as a tool to support educational management, favoring the adoption of more effective preventive measures.