Dropout at UnB under the machine learning lens: a dissertation predicts who is at risk
Edmilson Cosme da Silva defended at UnB’s PPCA a dissertation applying machine learning to predict academic dropout in UnB’s on-campus undergraduate programs.
Brasília, July 3, 2023
Undergraduate dropout is a collective pain that Brazilian universities still measure poorly. It is known to be high, to cost resources, to erode enrollment — but it rarely reaches the point where pedagogical intervention is applicable in time. The dissertation defended in July 2023 by Edmilson Cosme da Silva at UnB’s PPCA steps into that gap, proposing prediction instead of observation — under the advising of Prof. Dr. Sergio Antônio Andrade de Freitas .
The work takes UnB’s on-campus undergraduate programs as scope and uses machine learning to identify indicators associated with dropout. What makes the proposal relevant is not the technique itself — predictive models abound in the literature — but the effort to discuss which indicators, in the Brazilian context, yield signals early enough for coordination teams to act. A model that hits 90% once the student has already left is a curiosity; one that flags risk mid-semester is actionable.
By linking Learning Analytics to applied machine learning in educational policy, the work offers UnB — and similar institutions — input to redesign student support systems. It is the kind of contribution that must migrate from PDF to the coordinator’s dashboard to generate impact.
The research also led to the paper A Systematic Review of the Factors that Impact the Prediction of Retention and Dropout in Higher Education, published at the 56th Hawaii International Conference on System Sciences. The paper organizes the literature behind the predictive reading explored in the dissertation.
The defense took place on July 3, 2023 in Brasília. The examining committee was chaired by the advisor, Prof. Dr. Sergio Antônio Andrade de Freitas (FCTE/UnB), and composed of Prof. Dr. Rafael Ferreira Leite de Mello (UFRPE) and Prof. Dr. Andrea Felippe Cabello (FACE/UnB). Prof. Dr. Gladston Luiz da Silva, PPCA program coordinator, also attended the session.
The full text — written in Portuguese — is available at the UnB Institutional Repository.
About CEDIS: The Center for Studies, Development, and Innovation in Software (CEDIS), linked to the University of Brasília, researches and develops innovative software solutions with presence in Learning Analytics .