Retenção na educação superior: uma investigação baseada em mineração de dados na UFC - Campus de Crateús

Abstract

This study investigates student retention in higher education. It analyzes Brazilian educational policies, such as the Support Program for the Restructuring and Expansion Plans of Federal Universities (REUNI), the National Student Assistance Policy, and the Quota Law, which have fostered democratization, access, and student retention in higher education institutions. The study examines challenges related to completing undergraduate programs within the expected timeframe and student dropout. The objective was to identify factors associated with student retention in undergraduate programs at the Federal University of Ceará (UFC). Conducted at the UFC Crateús campus, the study found that the Graduation Success Rate remained below the institutional average over a ten-year period, with low graduation rates and high retention levels. The population consisted of 565 students admitted between 2016 and 2024, with active enrollment in the 2025.2 semester. This applied, descriptive, and exploratory study adopted a mixed-methods approach with methodological triangulation. Data were collected through document analysis and questionnaires. For data analysis, data mining techniques were employed using supervised machine learning methods to identify retention patterns. The algorithms applied included Decision Tree, Random Forest, K-Nearest Neighbors, Logistic Regression, and XGBoost, evaluated using Accuracy, Precision, Recall, and F1-score metrics. A cross-validation procedure (K = 10) combined with an 80/20 holdout split for training and testing was conducted. The Wilcoxon statistical test indicated no significant difference between the models (p > 0.05), and KNN was selected for presenting maximum Recall equal to 1.00, with Accuracy ≈ 0.60, Precision ≈ 0.57, and F1-score ≈ 0.72. However, the model does not yet demonstrate sufficient performance for immediate application, requiring further improvements before implementation. Nevertheless, the DEYSI system partially achieved its objective in the extraction, transformation, and loading stages. The results showed that 70.4% (398) of students were retained, while 29.5% (167) were in regular academic standing. Among the retained students, 74.0% (294) were classified as critical retention and 26.0% (104) as moderate retention. By field of study, the highest retention rate was observed in Exact and Earth Sciences (55.7%, 222 students), followed by Engineering (44.2%, 175 students). According to the students, the main factors associated with retention include failures in foundational courses, deficiencies in prior education, early choice of academic programs, rigid and extensive curricula with a high density of prerequisites, resistance to prerequisite flexibility, limited course offerings, and insufficient student assistance and financial aid. It is concluded that retention is a multifactorial phenomenon, requiring preventive and continuous institutional actions, strengthening of existing strategies, enhanced dialogue among academic stakeholders, and expansion of student assistance policies.