O impacto da autoeficácia na evasão de ingressantes do ensino superior particular brasileiro

• Ano de defesa: 2023
• Autor(a) principal: Cunha, Átila Simões da
• Orientador(a): Zambaldi, Felipe, Yoshinaga, Claudia Emiko
• Tipo de documento: Tese
• Tipo de acesso: Acesso aberto
• Idioma: por
• Palavras-chave em Português: Modelos preditivos; Aprendizado de máquina; Autoeficácia; Ensino superior; Evasão
• Palavras-chave em Inglês: Predictive models; Machine learning; Self-efficacy; Higher education; Dropout
• Link de acesso: https://hdl.handle.net/10438/33828

The recent expansion of Brazilian higher education led to an increase in the number of seats, above the demand growth, which, followed by significant dropout rates, ended up creating a significant number of idle seats, requiring practices from the educational manager to mitigate the impacts of this phenomenon. This dissertation deepens the understanding of the dropout phenomenon in higher education, thus creating conditions for the adoption of new university management practices capable of reducing idleness, by raising the retention rate and decreasing school dropout, resulting in improved performance and progress towards achieving target 12 of PNE. The literature review points to the existence of explanatory models of evasion, but the lack of models combining observable and latent variables, from the field of psychology, and considered to have the greatest impact on evasion, especially the variable ‘perceived selfefficacy’, is an important research gap. As a way of strengthening these studies, this dissertation proposes the following research problem: “What is the impact of freshmen’s perception of selfefficacy in private higher education on school dropout that occurs in the first semester of university life in Brazilian Private Higher Education Institutions?” The methodological approach was quantitative, based on cross-sectional analysis of secondary data using a sample containing more than 200,000 students for the period from 2017 to 2021, and a survey. Data were processed through multivariate statistical analysis techniques, factor analysis, and machine learning techniques. The conclusions showed the relevance of using machine learning techniques as predictive dropout methods, as well as the positive impact of self-efficacy in reducing dropout for students between 19 and 25 years old. The research results intend to support public policy makers and university managers for improving their knowledge of the dropout phenomenon, thus contributing to the adoption of new university management practices capable of reducing idleness, by raising the permanence rate and decreasing school evasion, leading to the improvement of the educational enterprise performance, and progress towards the achievement of goal 12 of the PNE, from the perspective of social contribution.