CLASSIFICAÇÃO E IDENTIFICAÇÃO DE PERGIS DE EVASÃO DO ENSINO MÉDIO INTEGRADO DO CAMPUS CERES DO IF GOIANO

Abstract

Despite all the changes and transformations that education has undergone to the present day, the country still lacks development, especially to resolve school dropout rates, which has consequences for society as a whole. Therefore, the goal here is to build a predictive model based on a data mining algorithm that best fits the conditions and data arising from the technical courses integrated into the high school on the Campus Ceres of the IF Goiano. For this, the KDD methodology is used to guide the entire data mining process. The sample was obtained from the database of the academic system and the selection process, both provided by the institution included in the research. The dataset had a total of 1,769 records which, after cleaning and transforming the data, amounts to 1,478 enrollments, from the years 2015 to 2020, and 28 attributes, with 158 of these enrollments being evasion ones. The Weka tool and algorithms that it provides were also used. To suggest the best model, it was necessary to compare the most used algorithms in the literature, which are decision tree, SVM, kNN, Naive Bayes, RNA, and logistic regression. After comparison, the Naive Bayes algorithm was considered the best predictor for the present dataset, as it delivers a low error rate (16.3%) about dropout students and good accuracy (89.6%).