<![CDATA[Machine learning approaches may establish a complex and non-linear relationship among input and response variables for the assessment of the Basic Education Development Index (IDEB) database and show indicators that may contribute to monitoring the quality of education. This paper uses extensive experimental databases from public schools, consisting of a case study in Brazil, to analyze data such as the physical and technological structure of schools and teacher profiles. The research proposes decision tree-based machine learning models for predictions of the best attributes to positively contribute to IDEB. It employs a newly developed SHapley Additive exPlanations (SHAP) approach to classify input variables, so to identify variables that impact the most the final model; a non-probabilistic sample was used, composed from three official databases of 450 schools, and 617 teachers. Results show that the number of computers per student, teachers’ service time, broadband internet access, investments in technology training for teachers, and computer labs in schools are the variables that have the greatest effect on IDEB. The model applied shows high prediction accuracy for test data (MSE = 0.2094 and R² = 0.8991). This article contributes to improving efficiency when monitoring parameters used to measure the quality of a teaching-learning process. Doi: 10.28991/ESJ-2022-SIED-020 Full Text: PDF]]>
Copyrights © 2022
