Processo decisório de gestão da evasão nos cursos de graduação da Universidade Federal do Sul da Bahia: contribuições da abordagem da mineração de dados educacionais

Abstract

Education is essential for democratic Public Administration, as it promotes equity, economic development, and social inclusion. However, despite advances towards universalization, complex challenges persist, and dropout is a major obstacle, demanding efficient Public Policies. This research aims to propose specific strategies for the management of academic dropout in undergraduate courses at the Federal University of Southern Bahia (UFSB) based on the use of the Educational Data Mining approach, delivering a Consulting Report, to address this problem and suggest practices to improve the management of dropout in the institution. The research method is based on a quantitative approach, of an applied nature and combined with explanatory exploratory research. Based on a bibliometric study, composed of the theoretical framework and bibliographic review with the application of the Theory of Consolidated Meta-Analytic Approach (TEMAC) associated with documentary research, secondary, quantitative data were investigated, with Logistic Regression and Decision Tree analysis applied to the case study. The initial sample is composed of secondary data, with 11,626 lines related to undergraduate courses and 502,403 related to disciplines, from undergraduate students at the Federal University of Southern Bahia when 23 variables were named. The sample was prepared by observing the normality curve and identifying missing data and outliers. The results indicate the Logistic Regression performed by course, highlighting as the main variables the CR (β 0.362) and Permanence (β 0.139) and by disciplines, the relevance that fell on the Final Average (β 1.759) and Total Workload (β 0.007). The research was deepened with Decision Tree modeling, presenting an accuracy of 86.7% and revealing that the most relevant factor in predicting dropout at UFSB is the length of Permanence, followed by Performance Coefficient (CR), Age, Campus, and Distance from residence to the university. The results highlight topics not found in other scientific literature of evasion. The purpose of the work was fully achieved with the preparation of the Technical-Technological Product, delimited in a Dashboard of the Evasion Observatory presented in the format of a Consulting Report containing intervention proposals and suggestions for application development.