Mineração de dados socioeconômicos e educacionais de discentes para predição de evasão e retenção escolar

Abstract

Dropout and school retention have always been topics addressed in Brazilian education. These issues impact not only students’ lives, their families, and the society they live in but also the budget of educational institutions. This is evidenced by the fact that the high dropout and retention rates represent a waste of public resources. In light of this, educational institutions must fulfill their role as educators and seek innovative approaches to allocate their resources effectively in combating dropout and retention. Educational Data Mining enables the understanding of factors that can enhance the educational proposal, as well as the prediction of students’ performance and the factors influencing learning. For a deeper understanding of the topic and the current state of the art, a Systematic Literature Review was conducted. This review allowed for the identification of the most widely used Artificial Intelligence algorithms, along with the associated data. As a result, this SLR played a fundamental role in improving the understanding and defining the requirements for the proposed software. Based on these characteristics, the goal of this work is to develop a tool that utilizes educational and socioeconomic data mining. Through the application of classification techniques, the tool aims to assist educational managers in addressing dropout and retention from the moment of enrollment until the start of classes. This tool was developed in Java, using the Weka library. In the experiment validation process, two distinct databases were used, resulting in an impressive accuracy rate of over 97% in both databases. To assess the software’s usability, the System Usability Scale questionnaire was administered, with two additional questions added to better understand potential difficulties in using the software within the school community. The questionnaire was conducted by education professionals in schools located in the region of the Pernambuco hinterlands. The results of this validation process provide valuable insights into the tool’s performance and usability, contributing to its ongoing evaluation and improvement.