Um estudo comparativo da análise dos erros nos descritores das avaliações em larga escala SPAECE e SAEB com aplicação da ciência de dados

Abstract

Data generation in the field of education has grown significantly in recent years, driven mainly by data from large-scale assessments such as SAEB. SAEB is a large-scale assessment applied to students in public and private schools throughout Brazil, with the aim of evaluating their level of learning. Given this context, it becomes crucial to adopt approaches that explore data analysis through Data Science, a field that provides us with various resources to extract crucial information from data sets, making it possible to extract valuable insights for the field of education. This work focuses on applying the FP-Growth algorithm to analyze SAEB data. FP-Growth is widely used in the field of Data Science (DS) to identify patterns of association. By applying the FP-Growth algorithm to SAEB data, the aim is to identify patterns of association related to the topics in which students faced difficulties during the assessment. The insights gained through this approach can provide valuable information for educators and others responsible for the area, allowing the identification of possible challenges faced by students that may have previously gone unnoticed, hidden by the mass of data. This work aims to contribute to the improvement of the educational process, allowing for more targeted and effective intervention in specific areas of difficulty, thus promoting substantial improvements in educational performance. To achieve this goal, we first outlined all the characteristics of the research, then adopted the DC process, which consists of six phases, starting with defining the objectives, then retrieving the data, then preparing, exploring, modeling and finally presenting the results. In the last phase of the DC process, we present the results, where we can analyze a ranking with the main difficulties faced by students in the Municipality of Fortaleza, in the subject of mathematics, in SAEB 2019. In addition, we then present the relationships between these main difficulties faced by the students, demonstrating issues that are often antecedents and consequents between the main errors presented by the students. After presenting the results of the SAEB 2019, we applied a comparative analysis with the results achieved by another study that used data from the SPAECE 2019, making it possible to verify the consistency between these two assessments. This approach is fundamental for analyzing students’ difficulties, since by identifying the relationships between problematic topics, educators can adjust teaching strategies, develop more personalized approaches and implement effective pedagogical interventions. In this way, analyzing the relationships between the most difficult subjects not only points to superficial deficiencies, but allows the problem to be tackled at its source