27February2026
10:00 Master's Defense fully remotely
Topic on
Decision Support Systems in School Management: Use of Explainable Machine Learning and Scenario Simulation for Data-Driven School Management
Student
Kamila de Almeida Benevides Costa
Advisor / Teacher
Marcos Medeiros Raimundo - Co-advisor: Germain García Zanabria
Brief summary
The evaluation of the quality of basic education in Brazil, based on large-scale instruments such as the Basic Education Assessment System (SAEB), generates a vast volume of data that presents significant interpretation challenges for school administrators. Although Machine Learning (ML) techniques offer great analytical potential to deal with this complexity, the opacity of traditional models ("black box") limits their practical adoption and confidence in decision-making. This dissertation proposes the development of a Decision Support System (DSS) in the form of an interactive dashboard, which integrates the predictive capacity of the XGBoost algorithm with Explainable Artificial Intelligence (XAI) techniques, specifically SHapley Additive exPlanations (SHAP) and Counterfactual Analysis. Using microdata from SAEB 2019, the methodology allowed the identification and quantification of the influence of intra- and extra-school factors on school performance. The results corroborate the literature by highlighting the strong impact of socioeconomic variables, but they go further by enabling the identification of "neighboring schools"—institutions with similar contexts but divergent performances—facilitating contextualized benchmarking. Furthermore, the developed counterfactual simulation tool allows managers to visualize hypothetical scenarios and quantify the potential impact of specific interventions, transforming complex diagnoses into actionable strategies for improving management and educational quality.
Examination Board
Headlines:
Marcos Medeiros Raimundo IC / UNICAMP
Rodolfo Jardim de Azevedo IC / UNICAMP
Celmar Guimaraes da Silva FT / UNICAMP
Substitutes:
Julio Cesar dos Reis IC / UNICAMP
Luis Gustavo Nonato ICMC / USP