Análise e discussão da predição de um caso de evasão universitária: uma abordagem baseada em data mining e machine learning

Abstract

Undergraduate courses in Brazil have high dropout rates, with a national average of 18% and, in the context of the state of Ceará, reaching 25% in on-campus courses. This phenomenon has significant impacts on both students and educational institutions, compromising the country’s educational development. Given this scenario, this study aims to carry out a data mining project on higher education student data, with a specific focus on the Computer Science program at the Quixadá Campus of the Federal University of Ceará (UFC), and to analyze and discuss the results of this process. Machine learning techniques were applied to data from academic records of former students to understand the factors associated with dropout and thus propose more effective mitigation strategies. For each Machine Learning algorithm tested, best practices in Data Mining were applied, such as Hyperparameter Tuning and Cross-Validation. Significant correlations between the existing columns, including the target attribute, were also demonstrated.